Deep convection, as used in meteorology, refers to the rapid ascent of air parcels in Earth’s troposphere driven by the buoyancy generated by phase change in water. Deep convection undergirds some of Earth’s most important and violent weather phenomena and is responsible for many aspects of the observed distribution of energy, momentum, and constituents (particularly water) in Earth’s atmosphere. Deep convection driven by buoyancy generated by the radiative heating of atmospheric dust may be similarly important in the atmosphere of Mars but lacks a systematic description. Here we propose a comprehensive framework for this phenomenon of dusty deep convection (DDC) that is supported by energetic calculations and observations of the vertical dust distribution and exemplary dusty deep convective structures within local, regional, and global dust storm activity. In this framework, DDC is distinct from a spectrum of weaker dusty convective activity because DDC originates from preexisting or concurrently forming mesoscale circulations that generate high surface dust fluxes, oppose large-scale horizontal advective–diffusive processes, and are thus able to maintain higher dust concentrations than typically simulated. DDC takes two distinctive forms. Mesoscale circulations that form near Mars’s highest volcanoes in dust storms of all scales can transport dust to the base of the upper atmosphere in as little as 2 h. In the second distinctive form, mesoscale circulations at low elevations within regional and global dust storm activity generate freely convecting streamers of dust that are sheared into the middle atmosphere over the diurnal cycle.
It has been proposed that hydrogen escape from Mars’s upper atmosphere is enhanced during dust storms, particularly the global dust events that affect Mars every few Mars years (Chaffin et al. 2014; Chaffin et al. 2017; Heavens et al. 2018). Under this proposal, dust storms transport water vapor from near Mars’s surface to the middle atmosphere (approximately 50–100 km above the surface), where water vapor is photolyzed into atomic hydrogen by extreme ultraviolet (EUV) radiation. The pathway between water vapor in the middle atmosphere and escaping hydrogen in the upper atmosphere has yet to be validated observationally (Jakosky et al. 2018), but up to tenfold increases in middle-atmospheric water vapor were observed during the two most recent global dust events in 2007 and 2018 (Fedorova et al. 2018; Heavens et al. 2018; Vandaele et al. 2019). Moreover, for global and some smaller dust events, there is synchrony between warm tropical temperatures at 25 km above the surface (a sign of atmospheric heating by dust above the boundary layer; Kass et al. 2016), vertical transport of dust to the middle atmosphere on time scales faster than a sol (a Martian day, which is 1.02 Earth days), and high amounts of water vapor being observed above the climatological hygropause (analogous to Earth’s tropopause) on Mars’s dayside (30–50 km above the surface).
Vandaele et al. (2019) cited four explanations for increased middle-atmospheric water vapor during planetary dust events: 1) solar radiation absorbed by dust heats the atmosphere and increases the saturation vapor pressure at higher altitudes; 2) atmospheric heating by dust increases equator–pole temperature contrast and strengthens the planetary-scale mean meridional circulation, enhancing mixing between the equator and the pole (and presumably vertical mixing in the ascending and descending branches of the circulation); 3) localized deep convection driven by the solar heating of dust (Rafkin 2012; Spiga et al. 2013); and 4) large-scale ascent of dust layers driven by the solar heating of dust (Daerden et al. 2015). Heavens et al. (2018) proposed that the correlative relationship of middle-atmospheric dust transport to with middle-atmospheric water vapor as well the speed of dust and water transport argued for localized deep convection being the enabling transport mechanism.
To paraphrase Houze (1997), “deep convection” is a term often used in meteorology of the Earth, yet the term is not defined in the Glossary of Meteorology, and its meaning and usage continually evolve. Individual investigators develop and apply empirical definitions appropriate to their methods and observational domains (e.g., Pettersen et al. 1945; Liu and Liu 2017). So, some discretion is appropriate when applying it to Mars.
Here we modify the definition of Houze (1997), so that convection is “the rapid, efficient, vigorous overturning of the atmosphere required to neutralize an unstable distribution of density” in order to be inclusive of moist convection and dry convection in the boundary layer as well as types of atmospheric convection unknown on Earth. Moist convection is thought to be of minimal importance in Mars’s thin, relatively cool, and predominantly CO2 atmosphere (Colaprete and Toon 2000). Nevertheless, Mars still has three particularly potent buoyancy sources for atmospheric convection: 1) the radiative heating/cooling of dust (e.g., Fuerstenau 2006; Heavens et al. 2011b; Spiga et al. 2013); 2) the radiative heating/cooling of water ice (Hinson and Wilson 2004; Spiga et al. 2017); and 3) latent heating/cooling resulting from the deposition/sublimation of CO2 (Colaprete et al. 2003; Määttänen et al. 2010; Hayne et al. 2012; Hayne et al. 2014).
We further distinguish deep convection from shallow convection based on vertical extent and impact on vertical transport. Warner et al. (1979, 1980) defined shallow convection as being primarily confined to the “moist layer” of uniformly mixed water vapor approximately corresponding to the planetary boundary layer (PBL), and deep convection as everything that penetrated higher. On Mars, shallow convection then would be everything above a height of up to 10 km (Hinson et al. 2008).
But further gradations are possible, depending on the transport and impact of interest. Vertical transport by moist deep convection in Earth’s tropics maintains a minimum in moist static energy in the middle of the troposphere, mainly by moistening the upper troposphere (Riehl and Malkus 1958). The strongest updrafts in thunderstorms on Earth penetrate the tropopause and moisten the stratosphere (Holton et al. 1995; Sherwood and Dessler 2000). And this transport is not limited to water but extends to trace gases, momentum, and energy, disturbing radiative and chemical equilibrium in the troposphere and beyond (e.g., Helfand 1979; Dickerson et al. 1987). The analogous convection at Mars would be the convection that mixes dust and water vapor into the middle atmosphere, which is called dusty deep convection (DDC) in line with the deeply penetrating convection simulated by Spiga et al. (2013).
DDC was first suspected by Gierasch and Goody (1973), who proposed Martian dust storms might resemble hurricanes and thus have deeply mixed dust concentrated near their centers with vertical velocities on the order of 0.1 m s−1. Furthermore, plume-like, cumuliform, and/or cellular organization of dust clouds long has been described in imagery (Kahn 1984; Määttänen et al. 2009; Kulowski et al. 2017), though some of this activity either was thought to be mechanically forced (Kahn 1984) or is now known to be confined to the PBL (Heavens 2017).
Interest in DDC resurged after Rafkin (2003) echoed by Strausberg et al. (2005) suggested that cumuliform dust clouds observed in the early stages of the 2001 global dust storm could be the result of deep convection. This proposal was buttressed by modeling of vertical transport related to katabatic/anabatic circulations (Rafkin et al. 2002; Michaels et al. 2006) and of tropical cyclone-like dynamics in Martian dust storms (Rafkin 2009).
The discovery of detached layers in the distributions of dust and water (local maxima in dust or water vapor mixing ratio above the PBL) in the Martian atmosphere (Heavens et al. 2011a,b; Guzewich et al. 2013; Maltagliati et al. 2011, 2013) and the unusual properties of detached dust layers near Mars’s highest volcanoes (Heavens et al. 2015) further strengthened the observational case for DDC. Just as Riehl and Malkus (1958) inferred the existence of deep moist convective clouds from the thermal structure and water vapor distribution of the tropical atmosphere, investigators inspired by Rafkin (2003) have inferred the existence of DDC from the vertical distribution of dust and water vapor on Mars. The discovery of widespread detached layering and cellular structure within a local dust storm (Määttänen et al. 2009) motivated the simulation of a deep convective dust cloud in an idealized local dust storm within a mesoscale model (Spiga et al. 2013). In the past year, a global climate model (GCM) parameterization of deep convective dust clouds has proven successful at reproducing some aspects of vertical dust layering in the Martian atmosphere (Wang et al. 2018).
The purpose of this study is twofold: first, to establish observationally that DDC on Mars is energetically and kinematically analogous to moist deep convection on Earth, and second, to provide an observational overview of deep convective dust clouds and associated structures in Martian dust storms. Indeed, these clouds/structures seem to encompass both localized deep convection and the ascent of large-scale dust layers cited by Vandaele et al. (2019).
Such a study is necessary for three reasons. First, DDC needs better description than in scattered references in the planetary science literature that focus on special cases, individual events, and connections to other atmospheric processes and/or other planetary atmospheres (e.g., Rafkin et al. 2002; Rafkin 2012; Spiga et al. 2013; Heavens et al. 2015, 2018; Heavens 2017). Second, improved profiling of the atmosphere at higher altitudes and opaque, dusty conditions by the Mars Climate Sounder (MCS) on board Mars Reconnaissance Orbiter (MRO; a key source of information about DDC in the past) now makes it possible to obtain temperature and aerosol opacity information on the margins of deep convective dust clouds in dust storms. Third, observations of the 2018 global dust storm by the Mars Atmosphere and Volatile Evolution (MAVEN) and ExoMars Trace Gas Orbiter (TGO) spacecraft likely include examples of DDC, so it is important to provide contextual information about DDC structures from MCS data and visible imagery in order to test hypotheses about the sources of middle-atmospheric water vapor.
The outline of the remainder of the paper is as follows. In section 2, we provide a brief primer on the Martian calendar and dust storm nomenclature and then describe the observations and previously developed analysis techniques used in the paper. In section 3, we provide an accounting of the energetics of dusty deep convection at Mars. The purposes of this section are to demonstrate the analogy between DDC at Mars and moist deep convection at Earth, motivate a new technique for assessing the intensity of DDC in the middle atmosphere, and provide interpretive context for observations in the following section. In section 4, we describe and analyze the diversity of structures associated with DDC in Martian dust storms. In section 5, we discuss these results and their significance for past and future observations and modeling. The entire study is summarized in section 6.
a. Martian time and dust storm classification
All references to Martian time use a calendar where Mars year (MY) 1 began on 11 April 1955 and time within a given year is expressed as areocentric longitude (Ls): the angle between Mars and the Sun relative to the northern vernal equinox (Clancy et al. 2000; Piqueux et al. 2015).
Global and regional dust storms are extended weather events rather than individual weather systems. To orient those interested in the recent history of Martian weather, we refer to them here by their MY of occurrence and type. Global (often called planet-encircling) storms are denoted P, such that the 2001 global dust storm (Strausberg et al. 2005; Cantor 2007) is 25P, the 2007 global dust storm (Wang and Richardson 2015) is 28P, and the global storm that began in early June of 2018 is 34P. Up to three regional dust storms associated with significant dust mixing above the boundary layer form each year during southern spring and summer and are named A, B, and C (Kass et al. 2016). In MY 29, a regional dust storm formed in northern summer (Smith 2009; Heavens et al. 2011a) and so by analogy with Kass et al. (2016), we will call it 29Z. Individual dust storms outside of these events will be called local dust storms, though it is common to call any dust storm smaller than 1.6 million km2 local in the midst of a regional or global dust event. The reader is referred to the references cited in this paragraph and Cantor et al. (2001) for further background on dust storm classification.
b. Mars Reconnaissance Orbiter Mars Climate Sounder data
The primary data presented here are 1) calibrated limb radiance observations in nine broadband channels ranging from the visible to the thermal infrared (McCleese et al. 2007) that can be referenced to altitude based on instrument pointing and 2) vertical temperature, dust opacity, and water ice opacity profiles (from the surface to 80 km) retrieved with version 5.2.4 of the retrieval algorithm from a selection of nadir/off-nadir and limb observations by MCS (Kleinböhl et al. 2009, 2011, 2017), which can be referenced to pressure or altitude. The radiance observations in the thermal infrared (primarily the dust-sensitive A5 channel centered at 463 cm−1) will be reported in brightness temperature. All of the radiance and version 5.2.4 retrieval data used here are currently archived in the NASA’s Planetary Data System (PDS; MCS 2018a,b). Individual profiles from 34P used to make cross sections (as opposed to zonal averages or vertical flux analyses) were retrieved using version 5.3.2, which uses information from far-infrared channels in which dust is less opaque. These retrievals will not be publicly archived, because the algorithm improvements will be incorporated into a new version of the retrieval algorithm that will be applied to the entire radiance dataset during 2019.
Observing nearly continuously since 24 September 2006 (Ls = 111° of MY 28), MCS has observed the limb forward (mostly) or backward (rarely) in the track of MRO. MRO’s polar orbit has drifted between approximately 1430 and 1530 local solar time (LST) at the equator on the dayside during the course of its mission (Zurek and Smrekar 2007). MCS can look to either side and so during dedicated campaigns, it has observed about 90 min earlier or later in local time at the equator than the MRO orbit (Kleinböhl et al. 2013). It also makes on-planet views in the off-nadir (and in the nadir during a portion of MY 28) at horizontal resolution (accounting for smearing by spacecraft motion) of ~6 km (Hayne et al. 2012).
The resolution of limb radiance observations is ~5 km in the vertical and ~8 km in the horizontal for individual channels perpendicular to the detector array (Hayne et al. 2012). However, the horizontal resolution of a limb observation in the direction of pointing is potentially 200 km or longer, depending on the limb opacity (Kleinböhl et al. 2009). So, structures less than 200 km in width typically will be underresolved. MCS limb observations are typically made in sets of eight, in which individual observations are nominally spaced by ~6 km. The average of the last six of these observations is used by the retrieval algorithm. The spacing between sets of limb observations varies but can be up to 400 km. When plotting, radiance observations and retrievals will be interpolated, but whitespace will be put over interpolations greater than 400 km. Limb observations are planned to track an elevation on the limb relative to a fixed potential surface and so may observe radiance at a tangent height below the local surface elevation.
MCS retrievals can be used to estimate the dust mass mixing ratio (MMR) and the water ice MMR (Heavens et al. 2010a, 2011a, 2018). Uncertainties in dust MMR are at the 10%–20% level, while water ice MMR is uncertain to a factor of 2 because of uncertainty in water ice particle size. Zonal averaging follows McCleese et al. (2010), while the calculation of the difference between the lapse rate and the dry convective lapse rate Γ follows Heavens et al. (2010b).
It is also possible to use dust MMR estimates and temperature retrievals on nearly coincident orbits about a day apart to estimate vertical dust fluxes, reported here as orbital averages (Heavens et al. 2018). These flux estimates are corrected for apparent vertical transport because of meridional advection. Vertical flux estimates are not corrected for the effects of zonal advection, because the algorithm relies on differencing the dust fields between closely spaced orbits separated by approximately a Martian day. Correcting for zonal advection would require incorporating information from neighboring orbits, which are separated by ~28° of longitude. In addition, the vertical dust fluxes presented here differ from those presented in Heavens et al. (2018), because they do not correct for sedimentation in cases where there is no dust MMR observed at a particular altitude and latitude during either orbit. This correction eliminates an “echo” effect, where the brief appearance of large amounts of dust in one orbit results in estimating an artificially large vertical dust flux during the succeeding orbit.
Elevation data and plotted altitude are relative to the areoid defined by the Mars Orbiter Laser Altimeter (MOLA) on board Mars Global Surveyor (MGS). The areoid is a gravitational equipotential surface chosen by convention and used on Mars like sea level is on Earth. Elevation data are taken from the MCS retrieval product or the 16-point-per-degree map made from MOLA observations (Smith et al. 2003) (Fig. 1). Areas commonly referred to in Martian dust storm studies are labeled in Fig. 1 as well.
c. Visible imagery
Daily Global Maps from observations by the Mars Color Imager (MARCI) on board MRO (Malin and Edgett 2001; Malin et al. 2008c; Bell et al. 2009) were used to provide context for some MCS observations, including to infer winds at the level at which the optical path from the satellite becomes opaque. MARCI imagery is calibrated, photometrically corrected and simple cylindrically projected at a resolution of 16 points per degree. The raw imagery is archived in the PDS (MARCI 2018).
3. The energetics of dusty deep convection
a. Shortwave heating of dust
The strong analogy between DDC in Mars’s atmosphere and moist deep convection in Earth’s atmosphere can be illustrated by comparing the potential heating rates and vertical velocities in DDC with those observed in deep convection on Earth. The relationship between dust MMR qD and density-scaled opacity is a function of the dust opacity dzτ scaled by the air density ρ multiplied by the bracketed conversion factor μd that is based on assumptions about the extinction coefficient Qext,obs at the observed wavelength, effective radius reff, and mass density of dust ρD (Heavens et al. 2011a):
The extinction of dust is relatively flat across the solar spectrum (Ockert-Bell et al. 1997), so for direct incidence, the specific heating rate JD is approximately
where εD is the absorption efficiency of solar radiation by dust in the parcel, Qext,sol is the average extinction coefficient over the solar spectrum, and Fsol is the incoming solar flux. The heating rate is JD divided by the specific heat capacity cp.
For example, if Eq. (2) is evaluated for these parameters [εD = 0.11; Qext,sol/ Qext,obs = 7.3; cp = 756 J kg−1; μD = 1.2 × 10−2 (m−2 kg−1)−1; Heavens et al. 2011b], JD is 6.6 × 10−5 W kg−1 ppm−1 (W m−2)−1, and the heating rate is 3.2 × 10−4 K h−1 ppm−1 (W m−2)−1. Note that cp is strongly temperature dependent. The value used here corresponds to a temperature of 219 K (Heavens et al. 2010b). The value of cp is 35% lower at 150 K. Note also that the efficiency εD is uncertain, but we would argue our choice of value is acceptably conservative. The value of 0.11 used above comes from Fuerstenau (2006) and is equivalent to the additive inverse of the single scattering albedo (1 − ω0) used by Ockert-Bell et al. (1997) (i.e., ω0 = 0.89). Use of this value presumes no further atmospheric absorption of photons initially scattered by dust, yet Fuerstenau (2006) allows for multiple scattering/additional absorption by admitting εD could be up to 50% more.
There are alternate estimates on both the low and high ends. Wolff et al. (2009) used observations of dust at a wide range of viewing geometries to derive ω0 = 0.94, implying a much lower εD. Yet Spiga et al. (2013) modeled shortwave heating rates of 24 K h−1 for a dust MMR of 48 ppm, which implies εD of 0.38 (assuming insolation of 450 W m−2)−1: (MCD 2018).
To obtain heating rates, the potential range of dust MMR can be estimated from surface observations during dusty weather. First, the average Martian dust devil for which opacity could be determined by Greeley et al. (2006) had a diameter of 11 m and an average visible opacity across its diameter of 0.14. Assuming an air density at the surface of 1.3 × 10−2 kg m−3 (Fuerstenau 2006), the dust MMR in a dust devil is 1600 ppm. Second, a developing global dust storm (34P) came close to or moved directly over the Mars Exploration Rover (MER) Opportunity on 10 June 2018, resulting in a visible optical depth measurement of 10.8 (Smith 2018). Rearrangement and integration of Eq. (1) shows that this optical depth corresponds to a well-mixed atmospheric profile with dust MMR of 120 ppm (i.e., 0.09 units of visible opacity per ppm of dust near 0-km elevation). If this dust was confined in the first scale height, surface dust MMR would be no more than 190 ppm. However, the rarity of dust devil activity near Opportunity (Greeley et al. 2010; Waller 2011) argues against the dust being lifted locally, so surface dust MMR in dust storms could be higher in lifting regions, perhaps approaching the values observed in dust devils.
Thus, taking a conservative value for peak incoming solar flux in the summer hemisphere tropics (400 W m−2) (Heavens et al. 2011b), heating rates in dust storms would be 3.2 W kg−1 or 15.4 K h−1, which compares well with the heating rates modeled by Spiga et al. (2013) (despite the higher absorption efficiency), while those in dust devils would be 42 W kg−1 or 200 K h−1. Thus, dust devil heating rates would be comparable to those in the eyewalls of severe tropical cyclones or in the cores of midlatitude squall lines on Earth (Zhang et al. 2002; Tao et al. 2006). The inferred surface dust MMR near Opportunity would correspond to a heating rate comparable to the average across squall lines (Tao et al. 2006). But these heating rates only are realized during peak daytime insolation. At night, dusty air will cool or heat in the infrared according to the contrast between the temperature of the dust and the effective emission temperature of the surface/lower atmosphere (Heavens et al. 2010a; Spiga et al. 2013).
b. High-altitude dust as a tracer of DDC vertical velocity
Shortwave heating of dust is mainly opposed by adiabatic expansion of the rising air parcel (Fuerstenau 2006; Rafkin 2009; Spiga et al. 2013) at a rate of −gw, where g is the gravitational acceleration and w is the vertical velocity. Thus, the approximate maximum upward velocity of a parcel neglecting overshoot wc can be derived from the balance between adiabatic cooling and shortwave dust heating:
This limit is wc = 0.85–11.3 m s−1 for the 3.2–42 W kg−1 or 15.4–200 K h−1 heating rate range given above. These upward velocities are within the range previously observed and/or modeled in dust devils and deep convection of dust storms. (Greeley et al. 2006; Heavens et al. 2011b; Spiga et al. 2013).
At high altitudes, the terminal velocity of dust, υt may approach the updraft velocity. Following Heavens et al. (2010a), the terminal velocity υt of spherical dust particles in the Martian atmosphere is estimated to be
where k is a proportionality constant of ~15 s−1 and r is the radius of the dust particle. Thus, a 1.06-μm particle (the effective radius assumed by the retrieval algorithm of MCS) will have υt > 0.85 m s−1 at a density of 1.9 × 10−5 kg m−3, which, following the ideal gas law and zonal average temperature data from MCS (McCleese et al. 2010) in northern summer, should occur at a pressure of 0.5 Pa or 60- km altitude. Therefore, dusty updrafts will begin to clear by sedimentation above 60 km, depending on the size of the dust and the velocity of the updraft, such that dust transport to 80 km would require an updraft velocity of 8 m s−1, close to the mean updraft velocity measured by radar in Oklahoma thunderstorms at 2-km scales (Giangrande et al. 2013).
The connection between υt and updraft velocity will be used to constrain minimum upward vertical velocities in Martian weather systems, though these calculations will use the retrieved temperature and pressure in each case to estimate density, which can increase significantly in the middle and upper atmosphere during global dust storm activity (Withers and Pratt 2013). Thus, we will speak of υt levels and layers because Eq. (4) relates such levels and layers to those formed by contours in density or specific volume.
After the convective plume collapses, the injected dust will be diluted by horizontal advection and sedimentation (Heavens et al. 2014) as well as segregate by size (Kahre et al. 2008). The dust layer may remain coherent enough to reheat the next day and recover some of the altitude lost during the night (Spiga et al. 2013; Daerden et al. 2015). Smaller and/or flatter particles will have lower υt than estimated. Caution must be exercised in estimating υt from older structures (where larger grain sizes may be depleted), which only can be distinguished from younger structures by looking at changes over time in closely spaced observations.
However, υt estimates of deep convective structures observed even on the same day they formed may be underestimated by a factor of 2. Observations during 25P suggest that dust in middle-atmospheric DDL could have an effective radius of up to 2 μm, similar to that observed at lower altitudes during non–dust storm conditions (Clancy et al. 2010). An effective radius of 2 μm was recently reported by Clancy et al. (2019) for dust in a DDL observed during early northern summer of MY 30 at 55-km altitude over Syria Planum (14°S, 104°W). Furthermore, if the particles are too small (~0.1 μm), MCS observations will not be sensitive to them (Kleinböhl et al. 2015). In addition, the higher the υt where dust is observed, the more likely it is that dust has fallen from a much higher level in υt and altitude prior to observation.
c. Role of water in DDC energetics
Malin et al. (2008c) reported an association in visible imagery between “very convective dust storms” and water ice clouds and suggested this was due to water ice nucleating on dust lofted to altitudes where the atmosphere was saturated with respect to water vapor. While water ice cloud formation may indicate convective intensity, water ice cloud formation is an agent of convective inhibition of DDC under most circumstances.
A dusty updraft where water ice is nucleating on dust will transition from radiative heating of dust (as the dust changes to a more reflective aerosol) to latent heating by water vapor being deposited as ice. [By analogy with subvisible cirrus in Earth’s atmosphere, shortwave heating of Mars water ice clouds is negligible relative to longwave heating, which we consider below (Bucholtz et al. 2010).] The minimum ice deposition time scale ηice is one in which an updraft cooling at the dry adiabatic lapse rate goes from some saturation vapor pressure with respect to ice esi to a saturation vapor pressure of 0:
where γd is the dry adiabatic lapse rate (~4.5 K km−1).
The upper bound for the latent heating rate Jice is then
where qvap is the initial water vapor mass mixing ratio of the parcel and Ldep is the latent heat of deposition for water (2.83 × 106 J kg−1) (Murphy and Koop 2005). This heating rate is an upper bound, because it ignores the effects on parcel temperature by radiative heating of dust during the transition to ice or any other factor.
At the low latitude site of the Mars Science Laboratory Curiosity, the frost point at the surface is ~190 K (Martinez et al. 2016), and qvap is 60 ppm (McConnochie et al. 2018). Using the surface frost point as an estimate of the cloud condensation temperature, we calculate that updrafts ascending at 0.85–11.3 m s−1 would have a maximum potential latent heating rate of 0.11–1.5 W kg−1 or 3%–4% of the shortwave dust heating rate that would balance adiabatic cooling (see the previous subsection). Thus, even if qvap were 100 ppm as observed in the middle atmosphere at high latitudes (Vandaele et al. 2019), latent heating would be a minor contributor to DDC energetics.
Nor will transitioning from longwave heating by dust to longwave heating by water ice make up for reduced shortwave dust heating. If the longwave dust heating calculation in Heavens et al. (2011a) is recalculated for a parcel condensing water ice at 190 K above a dust storm-cooled surface on the dayside tropics at 240 K (Guzewich et al. 2019), dust infrared heating will be 2.8 × 10−3 W kg−1 ppm−1, or 11% of the shortwave heating rate, whereas the daytime water ice infrared heating rate under similar conditions will be 3.9 × 10−3 W kg−1 ppm−1. So, a 100 ppm water ice cloud would have a radiative heating rate equivalent to 12% of the shortwave dust heating of a parcel with a dust MMR of 120 ppm, the estimated typical dust MMR for active lifting during dust storms.
d. Role of background aerosol and atmospheric stability
In the discussion above, the background atmosphere has been approximated as clear and isothermal; it is not. The significance of background opacity is intuitive. Any aerosol in the path between the Sun and dust lifted at the surface will attenuate solar radiation and reduce shortwave dust heating, inhibiting DDC (Spiga et al. 2013). Thus, the optically thick dust haze generated by a regional or global dust storm will inhibit DDC, as it does other dynamical processes related to dust lifting (Newman et al. 2002a,b).
The role of background atmospheric stability is more complex. When the background atmosphere rapidly cools with height, a dusty parcel will not require as much heating to remain positively buoyant but may lose heat from entrainment of cooler environmental air. However, if moving through a temperature inversion in the background atmosphere below its level of equilibrium, a dusty parcel is still warmer than the background atmosphere and losing heat by entrainment. But the parcel is also becoming less positively buoyant as its temperature contrast with the surrounding atmosphere is reduced. Thus, high atmospheric stability with respect to dry convection will inhibit DDC and low atmospheric stability with respect to dry convection will enhance it.
This effect can be quantified by modifying Eq. (3) for wc to take account of the environmental stability such that
where Γ is the difference between the environmental lapse rate and the dry adiabatic lapse rate γd of ~−4.5 K km−1.
Outside of global dust storm activity, daytime atmospheric stability (as measured by Γ) is typically low within the first two scale heights (pressure > 100 Pa, equivalent to 0–15-km altitude) but increases rapidly to high levels (Γ > |γd|) in the winter hemisphere extratropics because of a temperature inversion and less rapidly elsewhere (Figs. 2a–d). Atmospheric stability is moderate (Γ ≈ |γd|) throughout the middle atmosphere (pressure < 1–10 Pa), except above the inversion in the winter extratropics. Therefore, deep convective structures observed in the moderately or highly stable middle atmosphere may contain significantly lower vertical velocities than they could have had in the less stable lower atmosphere. In addition, dusty convection that does not reach the middle atmosphere still may contain vertical velocities similar to deep convection on Earth because of the low stability below 20-km altitude. Comparing the observed dust distribution to υt directly demonstrates there are high vertical velocities within dusty convective structures but may greatly underestimate the maximum vertical velocity during the lifetime of the updraft.
The stability structure outlined in the last paragraph held at the commencement of both 28P and 34P (Figs. 2e,g), but atmospheric stability between 100 and 1 Pa (~15–50 km) generally decreased during both global dust events (Figs. 2f,h). Atmospheric stability also decreased in the extratropical middle atmosphere during both global dust events (Figs. 2f,h). Atmospheric conditions were too optically thick for successful temperature retrieval, but past modeling and observations suggest that atmospheric stability near the surface likely increased (Newman et al. 2002a; Smith 2004).
4. Deep convective structures
a. The planetary mean perspective
Dusty deep convective structures will be identified (in part) from local inhomogeneities in the atmospheric distribution of dust, so it is necessary to explain how they are distinct from the mean dust distribution and its variability at the planetary scale. The dust distribution of the Martian atmosphere is often vertically inhomogeneous, such that zonal mean dust MMR has a local maximum at an altitude of ~15–35 km (10–100 Pa) in the tropics (Figs. 3a–d,h), a climatological feature known as the High Altitude Tropical Dust Maximum (HATDM) (Heavens et al. 2011b). Similar features in individual profiles or sets of profiles are called detached dust layers (DDL). At southern summer solstice (Figs. 3d,h), the HATDM varies minimally between day and night (Heavens et al. 2014) and can be reproduced in a GCM by parameterizing dusty deep convective clouds in local and regional dust storm activity (Wang et al. 2018). At other seasons, particularly during the first half of the year, the distribution is more diurnally variable and parameterizing dusty deep convective clouds is less successful at reproducing the observed vertical dust distribution (Heavens et al. 2014; Wang et al. 2018) (Figs. 3a,b,e,f). Particle scavenging and/or convective transport by water ice clouds may be the missing process (Navarro et al. 2014; Heavens et al. 2014; Spiga et al. 2017), but DDC associated with dust devils (Heavens et al. 2011b) or topographic circulations also has been considered (Michaels et al. 2006; Heavens et al. 2015; Wang et al. 2018). However, although layering in the zonal mean dust distribution is at least a partial consequence of DDC, dust is rarely found in atmosphere with υt > 0.1 m s−1, suggesting that vertical velocities at synoptic-planetary scales are less than this velocity. [Dust near the winter pole or at substantially higher altitude than the main region of dust in Fig. 3 is mostly CO2 ice (Heavens et al. 2011a)].
During global dust storms, zonal mean dust MMR increases dramatically and substantial amounts of dust MMR are observed at υt > 0.1 m s−1 or even up to 1 m s−1 (Figs. 4b,d,f,h). Zonal average dust MMR values are 30–60 ppm, corresponding to well-mixed visible column opacities of 2.7–5.4, close to what has been observed at the surface during global dust storms (Colburn et al. 1989; Lemmon et al. 2015). During 28P, substantial dust MMR was present in a region of high υt above the southern pole but at lower υt poleward at night (Fig. 4b) but uniformly south of 45°N during the day (Fig. 4f). During 34P, there was an HATDM of unusual magnitude but typical altitude both day and night, but dust did reach levels with higher υt on the dayside than on the nightside (up to 0.3 m s−1) (Figs. 4d,h).
That the transport of dust above υt > 0.1 m s−1 distinguishes global dust storm conditions is not just important for inferring updraft velocities. Dust transported to that level would fall a scale height (and thus a factor of e in υt) in around one Martian day. υt > 0.1–0.3 m s−1 is thus the lowest atmospheric layer where dust must be renewed each diurnal cycle, presumably by globally widespread DDC.
Analysis of vertical dust fluxes captures such renewal. Before 28P and 34P began, vertical dust fluxes were minimal above 40 km but then increased at all altitudes from 40 to 80 km on the dayside and between 40 and 70 km on the nightside during both storms (Figs. 5a–d). Stronger vertical transport during 28P than 34P was partly an artifact of a more inflated atmosphere during 28P, an effect that can be visualized by looking at dust transport relative to υt in Figs. 5a–d. There is also greater longitudinal inhomogeneity in dust MMR during 34P (visible in Figs. 5c,d as greater orbit–orbit “gappiness”). Both storms transported substantial dust to levels with υt between 1 and 3 m s−1 on the dayside and greatly enhanced transport of dust to levels with υt > 0.1 m s−1 on both the nightside and the dayside. Global dust storms thus contain regions of vertical velocity of up to 3 m s−1 at the scale of individual MCS retrievals. That substantial fluxes are observed at generally higher υt on the dayside than the nightside suggests that upward vertical transport is almost always stronger on the dayside than the nightside, which is consistent with the vertical dust fluxes being driven by a process strongly coupled to insolation.
b. Detached dust layers associated with dusty deep convection
DDL are observed that are high in both altitude and dust MMR and thus fit the extreme criterion of Heavens et al. (2015). Figures 6a–f shows some examples. Dust MMR in these DDL is up to 600 ppm and υt in the dustiest parts of the layer can be up to 3 m s−1 (Fig. 6b). Some DDL have been traced to local dust storms that form on or around some Martian volcanoes: volcanic local dust storms (Heavens et al. 2015). But these attributions sometimes cannot be confirmed by visible imagery (Heavens et al. 2015), because the dust may originate from lifting that ended prior to visible imaging. All DDL shown here were observed over or near the plateau of Tharsis Planum but not over the summits of Olympus Mons and the Tharsis Montes (Figs. 6 and 1) and so 500–1000 km from the nearest volcano. (Fig. 6e is an ambiguous case.) The nearest alternate source of dust to the example in Fig. 6a was Valles Marineris, just to the east (Fig. 1) (Malin et al. 2008a). However, in all other cases, the nearest nonvolcanic dust storm in visible imagery was 1000 km to the southeast in Solis Planum (Fig. 1) or more distant (Malin et al. 2008b, 2010, 2014a,b,c).
DDL therefore can have dust MMR between that observed in global dust storms and dust devils but often are observed at significant distance from volcanoes or dust storms imaged at lower elevations. Therefore, DDL seem to be horizontal outflow from vertically deeper structures rather than the result of convection directly below the DDL.
Extreme DDL also are observed in global and regional dust storm activity in areas outside of Tharsis. As in the examples in Figs. 7a–d, these DDL often have lower dust MMR and are observed at lower altitudes/υt than those observed near volcanic local dust storms (Figs. 6a–f). However, DDL in global storm activity can be up to 5000 km long in the meridional direction in which MCS generally observes, substantially longer than the DDL observed in Tharsis in northern spring and summer attributed to volcanic local dust storms. Note also a DDL observed at nearly zonal orientation in south polar day during the early stages of 28P (Fig. 7a). In some instances, two or more DDL can be identified; and the lowest/second highest layer is at too high an altitude to be the climatological detached dust layer at 15–35 km (Figs. 8a–d), which suggests that DDL can form in the same area on multiple, successive sols or possibly even multiple times per sol (Fig. 8d).
Thus, DDL seem to be outflow from “cores” of well-mixed dust that have visible opacity much greater than unity (dust MMR ≫ 11 ppm), likely contain updrafts with vertical velocities much greater than 0.1 m s−1, and can last multiple sols. At Earth, classification schemes distinguish higher-altitude, tower-like deep convective clouds from lower-altitude convective anvil clouds associated with wind shear and sedimentation of ice particles, though they share a genetic connection (e.g., Fu et al. 1990). We propose that cores and DDL are analogously related.
The cores postulated above can be observed, though with some difficulty because their high dust MMR can make them opaque to limb observations up to the top of MCS’s vertical range. Therefore, we identify them by comparing cross sections of successful retrievals with the observed radiance fields.
c. Vertically extended dusty cores
1) Volcanic local dust storms
Cores create notable high aspect ratio features in radiance observations of the limb originally named “limb castellations” (e.g., Heavens et al. 2015), because they looked like a warm castle surrounded by cold atmosphere. An extreme example of such a feature was observed over Arsia Mons (8°S, 120°W) around northern autumn equinox of MY 29 (Figs. 9a–h), but no dust storm was observed over or near Arsia Mons in visible imagery (Heavens et al. 2015). [A similar feature was observed synchronously with a visibly imaged volcanic local dust storm on Arsia Mons two weeks earlier (Heavens et al. 2015, 2018)]. The feature contrasted with the surrounding atmosphere to an altitude that may have exceeded the range of the detector array at ~90-km altitude (Figs. 9a–d,h). But the feature was not continuous and had a minimum in radiance at 30–50 km on its south side with a shape known as a loop, which arises from a cloud rising and setting relative to the horizon observed by MCS as MRO orbits (Sefton-Nash et al. 2013). The loop suggests that the base of the dust was near 50-km altitude rather than at the local surface elevation. While this core is exceptional, cores almost always can be recognized in the temperature-sensitive A1–A3 channels (Fig. 9a) as well as dust-sensitive A5 (Fig. 9b), suggesting that they are associated with significant anomalies of both dust and thermal emission relative to the surrounding environment. Peak solar flux over Arsia Mons on this sol was ~650 W m−2.
The feature was just transparent enough on its north side to allow a single retrieval between 45- and 80-km altitude. Retrieved temperatures were up to 190 K (at 65–70 km altitude), 30 K more than the surrounding atmosphere (Fig. 9e). dust MMR calculated from the retrieval was 3800 ppm at 75-km altitude (Fig. 9g), corresponding to a dust opacity ~4 × 10−3 km−1, at least two orders of magnitude greater than random and systematic biases and almost enough to obscure the limb and prevent sufficiently precise retrieval (e.g., Heavens et al. 2011a; Kleinböhl et al. 2015). (It can be shown that the visible optical depth will be of order unity around 15 km below this level.) Moreover, the layer of maximum dust MMR was distinct from the layer of maximum water ice MMR (Figs. 9f,g), so the retrieval algorithm was not affected by spectroscopic ambiguity between ice and dust in A4 (e.g., Kleinböhl et al. 2009; Heavens et al. 2015).
High dust MMR is also suggested indirectly by the altitude of the dust and the magnitude of its associated thermal anomaly. The retrieved dust was at υt ≫ 1 m s−1 and top of the core was at υt > 10 m s−1, implying upward vertical velocities may have been > 10 m s−1, fast enough to transport dust through the inferred depth of the cloud (70 km) in 2 h. The heating of a parcel with dust MMR > 850 ppm would be necessary to overcome adiabatic cooling at that vertical velocity. The temperature retrieval is almost entirely independent of the dust retrieval (Kleinböhl et al. 2009). Therefore, the 30-K temperature contrast with the surroundings allows an independent estimate of the heating rate. The radiative relaxation time of CO2 for 10-km thermal perturbations at this altitude and temperature is 0.6 h (Eckermann et al. 2011), so the heating rate must have been at least 50 K h−1 to compensate for radiative cooling of the gas alone. If upward vertical velocities were 10 m s−1, the net cooling rate was an additional 180 K h−1, which suggests dust MMR must have been at least 1100 ppm to maintain the anomalously high temperatures retrieved. And this calculation neglects the reduction of radiative cooling/scavenging of dust particles implied by the narrow water ice layer of 100 ppm below the dust layer (Figs. 9f,g).
The core in Fig. 9 was unusually deep and dusty. More typical are cores observed over Olympus Mons (18°N, 134°W) and Arsia Mons (Figs. 10a–h), in which dust MMR on the order of 100 ppm is retrieved at υt of 1–3 m s−1. Note, however, that retrievals are mainly successful only on the edges of core structures and that significant radiance is observed at atmospheric levels corresponding to υt > 3 m s−1.
Core morphologies notably differ between the two volcanoes. Cores over Olympus Mons tend to be bilobate. The core in Fig. 10a was observed while the detector array was looking in-track just to the west of Olympus Mons, while the core in Fig. 10c was observed while the detector array was looking cross-track at Olympus Mons (and thus about 90 min later in local time). Each core looks bilobate, but the core observed at later local time (Fig. 10c) is wider and flatter. Cores observed over Arsia Mons (Figs. 9 and 10b,d) tend to be narrower, pointed, and contain loops, as if multiple discrete dust layers are observed.
These differences may stem from contrasts in the circulation/morphology of local dust storms over the two volcanoes. Visible imagery frequency captures spiral-shaped dust clouds over Arsia Mons around northern autumnal equinox (Malin et al. 2008c). [This is the same season in which the cores over Arsia Mons shown in Figs. 9 and 10 and in Heavens et al. (2015, 2018) were observed.] If vertical motions are reasonably strong, such a circulation might generate a corkscrew-shaped dust cloud that would look like a series of discrete cloud layers in limb observations. Over Olympus Mons, lobate features trending toward an extended layer imply strongly divergent meridional flow at the top of the storm. However, strong divergence was simulated by Rafkin et al. (2002) at the top of Arsia Mons spiral dust clouds but not specifically reported in analogous simulations of mesoscale circulation over Olympus Mons (Michaels et al. 2006).
Some cores observed in eastern Tharsis may not be centered on a volcano. During MY 28, three limb castellations (and thus cores) were reported by Heavens et al. (2015) near Noctis Labyrinthus (7°S, 102°W), an area ~1200 km east of Arsia Mons (Fig. 1). The first two cores were relatively low in altitude (50 km) and observed just to the west or directly over local dust storm activity near Noctis Labyrinthus early in the MARCI mission (Figs. 11a,b,e,f). The third possibly extended above the MCS vertical range (85-km altitude) and occurred synchronously with a local dust storm in Noctis Labyrinthus (Figs. 11d,h). In the intervening period, a core was observed on Arsia Mons synchronously with a local dust storm in Noctis Labyrinthus (Figs. 11c,g). Visible imaging strongly suggests the prevailing direction of advection in the southern tropics (at a level of 20–30 km under these (water ice) cloudy, dusty conditions) is to the south or southwest. In that case, the cores observed in Figs. 11a,b,d would be centered to the northeast of the ground track, that is, in Noctis Labyrinthus. However, the core observed in Figs. 11c,g suggests that a deep convectively active local dust storm on Arsia Mons can occur simultaneously with a local dust storm in Noctis Labyrinthus that is visibly indistinguishable from deep convectively active storms observed during the same few weeks. However, Noctis Labyrinthus and Arsia Mons are separated by ~20° in longitude, so they cannot be observed by two successive in-track orbits, which are separated by ~28° of longitude. (No core was observed in the orbit just to the east of Noctis Labyrinthus at Ls = 129.5°).
Cross-track observations confirm that local dust storm activity near Noctis Labyrinthus can be deep convective. During late fall of MY 30, one or more cores were observed near Noctis Labyrinthus (Figs. 12a–j), particularly in Fig. 12c, whose ground track passed directly over a local dust storm in Noctis Labyrinthus. Nearly synchronous cross-track observations over Arsia Mons were interrupted (likely by a spacecraft maneuver required by another instrument) (Fig. 12b), so the presence of a core over Arsia Mons cannot be verified by direct observation of the summit (Fig. 12b). Visible imagery suggests advection in the southern tropics was to the south (Fig. 12e). If a dusty core were centered on Arsia Mons but broad enough to be observed by tracks 3, 5, and 7 in Figs. 13a–c, it at least should have been observed by track 8. To the contrary, cores only were observed in tracks 3, 5, and 7 at 10°–20°S, the only views that look toward the storm in Noctis Labyrinthus (Fig. 13d). Therefore, the storm in Noctis Labyrinthus contains a core with similar characteristics to those observed over the volcanoes to its west.
2) Regional and global dust storms
During the early stages of global and regional dust storm activity, structures that resemble volcanic local dust storm cores are observed in areas other than on or near the Tharsis Montes, consistent with the evidence from the planetary mean dust distribution that DDC becomes more widespread during global and regional dust storms. In the first few days of 28P, two cores of this type were observed in the same area of the southern midlatitudes on two successive sols (Figs. 14a,b,e,f). Another core was observed in the southern midlatitudes during the regional dust storm C (Figs. 14c,g). During the early days of 34P, a core was observed on at least two successive sols in the southern tropics (e.g., Figs. 14d,h).
And though the features in Figs. 14a–h are exceptional for limb castellations in areas other than the Tharsis Montes, they are mostly unexceptional compared with those attributed to volcanic local dust storms. The dayside examples are not associated with dust MMR greater than 160 ppm and do not extend above υt of 1 m s−1 (Figs. 14a–c,e–g). The 34P example is particularly interesting (Figs. 14d,h). It consists of an area (up to 25-km altitude) of >200-K brightness temperature over the southern tropics with a local minimum near the equator from which emerges a bilobate DDL, which vaguely resembles a decaying volcanic local dust storm over Olympus Mons. However, there is no mountain or ridge here, just a canyon and low plateau at 1-km altitude or less. The minimum in brightness temperature corresponds to a similar minimum in off-nadir A5 brightness temperature, which is roughly 30 K lower than outside the minimum in limb brightness temperature. And nightside cross-track imagery 15 h earlier shows a core at this location up to 75-km altitude, which would suggest DDC was intense here on the previous sol.
As global dust storms develop, cores become harder to identify as the lofting of dust to higher altitudes or atmospheric heating reduces the radiance contrast of areas of higher and more deeply vertically mixing dust with the surrounding atmosphere. Cores still may be identifiable as breaks in retrieval coverage that have DDLs with high dust MMR retrieved on their margins, as in Figs. 14e–g.
3) Dusty deep convection in Tharsis during global dust storms
In Tharsis, however, DDC structures are so dusty and reach so high in altitude that they can be observed throughout the course of global dust storm activity. An example about three weeks after the start of 28P is shown in Figs. 15a–j. In the first orbit shown, dust was well mixed to 70 km with dust MMR < 100 ppm (Figs. 15a,f). During the subsequent orbit, two narrow features were observed up to 90-km altitude (Figs. 15b,g). In the following orbit, the radiance signature is bilobate like a volcanic local dust storm core over Olympus Mons, though it is centered 2500 km to the south of it (Figs. 15c,h). These features are consistent with vertical velocities ≫ 1 m s−1. In the next orbit, the core has mostly disappeared, but a layer of dust MMR up to 300 ppm is retrieved up to 80 km and seems to have a higher-altitude radiance signature (Figs. 15d,i). This feature may be a DDL >3000 km long originating from one or more cores similar to cores observed in volcanic local dust storms. Visible imagery shows a plume of dust being advected southwestward from Arsia Mons, but there is no positive proof that this is the center of the core.
About one week into 34P, dust lifting expanded close to the Tharsis Montes and DDC was observed. On two subsequent sols, a core was observed in the forward limb in one orbit and in the right limb on the subsequent orbit (Figs. 16a–j) but not in other orbits/geometries. This core exceeded 90 km and the detector array in altitude. Vertical velocities in the core therefore exceeded 3 m s−1. Retrieval of dust MMR of 600 ppm on the margin of the cores would be consistent with vertical velocities exceeding 4 m s−1 (Fig. 16a). The geometry of the observations suggests the core was centered near 15°S, 100°W, close to cumuliform dust clouds stretching southeastward 600 km from Arsia Mons (Fig. 16e). On the next sol, the core seems to have been centered near 0°, 100°W and close to bright dust clouds in visible imagery (Fig. 16j), though any relationship between dust clouds observed on the successive sols is necessarily speculative.
These examples are synchronous with enhanced dayside dust flux at υt = 0.3–1 m s−1 (Figs. 5c,d). In the case of 28P, the example DDC over Tharsis commences the main period of deepest vertical dust mixing, while the 34P example only commences a short, initial period of vertical mixing. However, further investigation of the 34P storm has suggested that cores over the Tharsis Montes are partly responsible for the later, longer period of deepest vertical dust mixing. Full consideration of DDC, vertical mixing, and their evolution during 34P is the subject of an ongoing, separate investigation that builds upon this work but is beyond its scope.
If all cores produced identifiable limb castellations, cores in global and regional dust storm activity would be two orders of magnitude less frequent than expected from the relative occurrence frequency of DDLs and cores in volcanic local dust storms, which challenges the idea that DDLs largely originate from cores. Heavens et al. (2015) found that the vast majority (21 259) of retrievals containing DDL with dust MMR > 47 ppm and at altitudes > 50 km observed during MY 28–32 occurred during 28P. During 28P, nine limb castellations were observed (three of which were in Tharsis). In contrast, 225 retrievals containing DDLs meeting the same criteria and 24 limb castellations were observed in and around Tharsis outside of regional and global dust storm periods. The contrast in zonal mean dust MMR (Figs. 4a–h) or middle-atmospheric dust fluxes (Heavens et al. 2018) between periods with and without global dust storm activity poses a similar quandary.
The resolution to this quandary are structures we will call streamers, in which areas of high dust MMR appear traceable from orbit to orbit but rise in altitude/υt while gradually decreasing in dust MMR (Figs. 17a–c). From an observational standpoint, as MCS looks progressively farther and farther westward on the day side, an area of high dust MMR initially observed at low altitudes may evolve into a DDL (Figs. 17a,c) or at least a region of deeper vertical mixing at later times (Fig. 17b).
One possible interpretation is that these DDL may originate from an unobserved convectively active region at similar or higher altitude. However, a streamer observed during the 31A regional dust storm suggests otherwise. The first orbit in time/easternmost in space observed the eastern margin of the developing storm (Fig. 18b), which may be advecting dust to the north (Figs. 18a). The second orbit observed the western margin of the storm, where again wavy, filamentous dusty air masses may be ascending northward. However, a DDL is observed at much higher υt over the tropics during the third orbit. This orbit is ~1500 km to the west of the regional storm and crosses mostly clear conditions in visible imagery, except perhaps for some haze in the northern tropics to the east of the track (Figs. 18a,b). Two local dust storms (one frontal) are observed in the northern midlatitudes to the west (Fig. 18b). But the fourth orbit’s observations suggest the westernmost of these storms is not the source of the DDL. We therefore infer that the regional dust storm is the source of the DDL but that the DDL has reached much higher υt than the outflow in the immediate vicinity of the storm and thus is unlikely to originate directly from some unobserved core at similar υt to the DDL in the third orbit. Part of this inference is based on the water ice clouds that surround the regional dust storm, which are climatologically unusual (Wang and Ingersoll 2002) and have been proposed to indicate highly convective dust storms (Malin et al. 2008c).
If the regional storm is the source, the convective ascent of the dust is most likely simultaneous with the advection. For if the DDL were dust advected from deeper vertical mixing than observed over the regional dust storm by the first or second orbits, the dust would need to be advected faster than the MCS observational ground track moves westward over Mars (~240 m s−1 at the equator), which is much faster than the zonal mean easterlies simulated at this season and altitude (~30 m s−1 at the equator) (Forget et al. 1999). But the actual winds are probably significantly faster, because the modeled winds would imply the dust in the DDL came from the area observed by the second orbit about 0330 LST. If the winds were twice as fast, the DDL must have originated from the storm at a more plausible time of 1100 LST. With dust MMR at the 100-ppm level, vertical velocities in the 0.1–1 m s−1 range would be expected, enabling a change in altitude of 2.5–25 km during advection and also enabling most of the mixing between the lower and middle atmosphere on diurnal time scales necessary to produce the contrast in the vertical dust distribution between global dust storm conditions and otherwise. This mechanism is functionally equivalent to the solar escalator mechanism discussed by Spiga et al. (2013) and named by Daerden et al. (2015) by analogy with a mechanism identified in wildfire smoke plumes on Earth by de Laat et al. (2012).
Thus, on the basis of observations, we conclude that low altitude, hard-to-observe cores within regional and global dust storm activity contribute to vertical mixing of dust into the middle atmosphere by generating streamers that are transported horizontally by the global circulation and vertically by DDC. Streamers thus are DDLs but are distinguished from other DDLs by rising to altitudes significantly higher than their dusty core circulations of origin.
The energetic arguments presented in section 3 suggest that a parcel of Martian air with dust MMR strongly contrasting with that of the surrounding environment should freely convect, though free convection will be conditional on insolation, local stability, water ice condensation, and likely other factors. Observations confirm this idea by showing that parcels with dust MMR > 1000 ppm may reach the upper atmosphere (>90 km) and that parcels with dust MMR on the order of 100 ppm will penetrate deep into the middle atmosphere (65 km). Extreme vertical velocities in global dust storm activity downwind of the Tharsis Montes have been inferred before (Clancy et al. 2010), but their full significance has not been realized.
In addition, these observations motivate refinements to modeling, particularly the detection of large thermal anomalies associated with the dust anomalies (e.g., Figs. 9a,e). This warm air increases overshooting but also enhances radiative cooling and lowers the altitude of radiative equilibrium. These thermal anomalies can be much larger than modeled (30–40 vs 5 K) (Spiga et al. 2013). However, mesoscale modeling generally has treated circulations with dust MMR < 100 ppm (Rafkin et al. 2002; Rafkin 2009; Spiga et al. 2013), though sometimes with higher dust absorption efficiency than our energetic calculations assumed. That larger values of dust MMR exist in mesoscale circulations is surprising but is as much indicative of the limited phase space of dust storm activity explored by mesoscale modeling as it is of potential model biases.
Nevertheless, the examples presented here are exceptional. The persistence of a climatological detached dust layer at 15–35 km suggests dusty convective structures with much lower equilibrium levels than in DDC exist. In addition, there is convective mixing of dust below the middle atmosphere but distinct from the climatological layer at 15–35 km that is more widespread in southern spring and summer than in northern spring and summer (Heavens et al. 2015, 2018). Furthermore, “puffy” cumuliform dust clouds seem to be widely observed in low-elevation areas outside of regional and global dust storm activity (Guzewich et al. 2015; Kulowski et al. 2017). Therefore, a large population of shallower and smaller cores than resolvable by MCS observations likely exists in local dust storms.
The question that arises is what distinguishes the dusty convection that produces the DDLs in the lower atmosphere from dusty convection which reaches the middle atmosphere. One part of the answer is that the climatological detached dust layer at 15–35 km caps a region of low stability (Figs. 2a–d), so it is possible for less dusty parcels to generate dusty convection below the climatological detached dust layer than above it.
Another part of the answer is hinted by middle- and upper-atmosphere-penetrating DDC associated with the Tharsis Montes. Modeling demonstrates that anabatic flow up the slopes of these volcanoes acts as a natural center of convergence/cyclonic circulation (Rafkin et al. 2002), which simultaneously generates high wind stresses over the area surrounding the volcano but concentrates the dust at the scale of the caldera. This is not true for other sources of lifted dust. Dust lifted by dust devils will tend to be diluted quickly by less dusty environmental air or be inhibited from heating because of attenuation of sunlight by dust higher in the dust devil column. Dust lifted by straight line winds in a frontal dust storm will advect or diffuse away. A strong, convergent, mesoscale circulation is necessary to lift dust and keep it concentrated at the levels necessary for vigorous free convection. One limit to dusty convection originating from such circulations is reduced heating of the atmosphere lower in the atmosphere because of attenuation of sunlight by dust higher in the atmosphere. This effect, of course, is weaker at higher altitudes for the same value of dust MMR. Another possible limit is water ice cloud formation, which is common over the Tharsis Montes and nearby areas throughout much of the year (Wang and Ingersoll 2002).
This study has found hints of DDC-producing circulations that are not on top of high volcanoes. In areas where DDC seems to occur in the same small area, for example, near Noctis Labyrinthus (Figs. 11–13), dust availability and/or a quirk of local topography such as slope winds along the fractured terrain of Noctis Labyrinthus (Michaels et al. 2006) may stimulate the occasional development of such a circulation. In regional and global dust storm activity (e.g., Figures 18b and 16e,j), such circulations may form in low-elevation areas because of the dynamical forcing of the wider storm. But the more dust spreads, the more weakly such circulations will contrast in temperature and dust MMR with the surrounding environment. This effect certainly makes dusty convective structures harder to observe and likely reduces dusty convective structure formation in the lower atmosphere generally. However, widespread dustiness in the lower atmosphere also reduces dust loss by entrainment of any structures that do form, enabling dust MMR (and heating rates) to be high enough to form the structures we have named streamers.
Two other factors may support streamer formation. First, there is reduced stability in the tropics between 20 and 50 km and in the middle-atmospheric extratropics during global dust storm activity (Figs. 2e–h). Thus, a dusty air mass that makes it out of the dark, stable atmospheric conditions near the surface will experience less convective inhibition. Second, warming/inflation of the background atmosphere by dust heating inhibits or at least raises the altitude of water ice cloud formation (Vandaele et al. 2019), which limits inhibition of DDC by water ice cloud formation within the dusty updraft.
Understanding the dynamics of these circulations inside and outside of large-scale dust storm activity should be of broad interest beyond meteorology. Variations in the form and intensity of DDC (including streamer formation) are critical for understanding water vapor transport to the middle atmosphere during dust storms (Heavens et al. 2018; Vandaele et al. 2019).
What these observations do not settle is whether dusty deep convective clouds and any underlying mesoscale circulations are analogous to mesocyclones or tropical cyclones on the Earth. Like tropical cyclones, Martian dusty cores are warmer than their surroundings aloft and persist multiple diurnal cycles. However, the core in Fig. 14d is colder than its surroundings up to 25 km; the Noctis Labyrinthus storms look like clusters of cumulus clouds forming into a long anvil rather than a vortex. DDC cores may be diverse in structure and/or lack good Earth analogs.
Left unsettled, too, is the significance of DDC for any larger-scale circulations in which it may be embedded. DDC may deepen circulations, increase their thermodynamic efficiency, and therefore generate stronger winds at the surface by the mechanism of Emanuel (1986) or simply by mixing middle-/upper-atmospheric winds to the surface. At the same time, dust mixed by DDC warms the atmosphere, eventually inhibiting these effects. One key parameter for this phenomenon is the advection of dust into the upper-level winds by DDC structures, which controls how fast dust spreads in the middle and upper atmosphere in opposition to sedimentation, as has been explored by past modeling (e.g., Haberle et al. 1982; Newman et al. 2002b; Spiga et al. 2013).
Finally, our analysis likely underestimates dust MMR and vertical velocities, because limb observations underresolve the 10–100-km scales of the mesoscale structures in visible imagery and the amount of atmosphere sampled by limb observations is a few percent on any given day, such that many cores may not be observed (Heavens et al. 2015). Further observational advances in understanding DDC will require instrumentation that can provide visible imaging throughout the course of a sol, vertical profiling information at horizontal scales closer to that of visible imagery, and/or high horizontal resolution imaging of middle-atmospheric temperatures, as may be possible from nadir microwave observations (e.g., Muhleman and Clancy 1995).
Here, we have developed an observationally motivated and justified interpretive framework for DDC, a form of atmospheric convection in Mars’s atmosphere analogous in energetics, kinematics, and impact upon vertical mixing to moist deep convection in Earth’s atmosphere. Our central, guiding result is that >70-km-high convective towers with vertical velocities exceeding 10 m s−1 can form over high Martian volcanoes because dust concentrations within them are high enough to generate radiative heating rates comparable to latent heating rates in midlatitude severe thunderstorms or severe tropical cyclone eyewalls on Earth. We conclude that the extreme dust content of these convective structures is made possible by intense dust lifting and confinement of dust by orographically forced mesoscale circulations.
DDC is not just limited to the volcanoes. Widespread, diurnally repeating convective activity mixing dust into the middle atmosphere is a distinctive aspect of the meteorology of global dust storms relative to all other weather conditions, suggesting mesoscale structures like those proposed to occur on the volcanoes are embedded within regional and global dust storm activity. The inhomogeneity of the vertical dust distribution outside of large dust storms and observations of cumulus-like dust clouds in visible imagery may imply DDC is the intense tail of a wide distribution of dusty convection in the lower atmosphere. Mars therefore presents an entirely new area of mesoscale convective dynamics with major implications on Mars for vertical mixing, atmospheric evolution, and forecasting dust storms; and on Earth, for developing robust general theories of mesoscale convective phenomena.
This work was supported by the NASA Mars Data Analysis, Solar System Workings, and Nexus for Exoplanet System Science programs (NNX14AM32G, NNX15AI33G, and NNX15AE05G). Work at the Jet Propulsion Laboratory, California Institute of Technology is performed under contract with NASA. N.G. Heavens thanks A. Kleinböhl, D. J. McCleese, J. T. Schofield, L. J. Steele, M. I. Richardson, and C. E. Newman for useful discussions. We also thank three anonymous reviewers for encouraging and helpful reviews.