Textured Dust Storm Activity in NE Amazonis–SW Arcadia, Mars: Phenomenology and Dynamical Interpretation

Abstract

Dust storms are Mars’s most notable meteorological phenomenon, but many aspects of their structure and dynamics remain mysterious. The cloud-top appearance of dust storms in visible imagery varies on a continuum between diffuse/hazy and textured. Textured storms contain cellular structure and/or banding, which is thought to indicate active lifting within the storm. Some textured dust storms may contain the deep convection that generates the detached dust layers observed high in Mars’s atmosphere. This study focuses on textured, local dust storms in a limited area within NE Amazonis and SW Arcadia Planitiae (25°-40° N,155°-165° W) using collocated observations by instruments on board the Mars Global Surveyor (MGS) and Mars Reconnaissance Orbiter (MRO) satellites. In northern fall and winter, this area frequently experiences dust storms with a previously unreported ruffled texture that resembles wide, mixed-layer rolls in the Earth’s atmosphere, a resemblance that is supported by high resolution active sounding and passive radiometry in both the near and thermal infrared. These storms are mostly confined within the atmospheric boundary layer and are rarely sources of detached dust layers. The climatology and structure of these storms is thus consistent with an underlying driver of cold air advection events related to the passage of strong baroclinic waves. While the properties of the studied region may be ideal for detecting these structures and processes, the dynamics here are likely relevant to dust storm activity elsewhere on Mars.

1. Introduction

Dust storms are Mars’s most notable meteorological phenomena. First observed by telescopic observers as obscured albedo features and then as “yellow clouds”, their equivalence to Earth’s dust storms was confirmed from the first spacecraft observations (Gifford 1964; Martin and Zurek 1993). The vast majority of storms are local: storms do not last from one sol (Martian day) to the next and have areas < 1.6 × 10⁶ km² (Cantor et al. 2001). Local storm frequency decreases with increasing area according to a power law until an area of 1.6×10⁶ km², beyond which larger storms are more common than would be expected from the power law (Cantor et al. 2001). These regional storms (1–2% of storms) last for a few sols (Cantor et al. 2001). In rare cases, global dust storms enshroud most of the surface area of the planet in a haze of dust.

Observations of the global dust storm in 2001 suggested that global storms are not associated with a single center of circulation and/or frontal boundary. Instead, they consist of multiple local/regional storms that form along the advancing dust haze itself or as the result of global teleconnections associated with the high loading of dust (Strausberg et al. 2005; Cantor 2007). Regional dust storms can form from merging local storms, but are often associated with discrete weather systems, particularly fronts (Cantor et al. 2001; Cantor 2007; Wang and Richardson 2015).

Global storms loft dust up to 75 km above the surface of the planet (Conrath 1975; Cantor 2007; Clancy et al. 2010; Heavens et al. 2015). This dust strongly heats the atmosphere while preventing sunlight from reaching the surface, which results in strong near-surface cooling during the day but warming (due to a dust-driven greenhouse effect) at night (Gurwell et al. 2005; Strausberg et al. 2005). Most of the dust forms a well-mixed haze, but some forms structures detached from the main haze, known as detached dust layers (DDLs). DDLs are also observed in Mars’s atmosphere outside of regional and global dust storm activity, but the dustiest and highest altitude DDLs are predominantly associated with global and some regional dust storms (Clancy et al. 2010; Heavens et al. 2011a,b, 2014, 2015).

The circulation within local storms is not as well-characterized. Modeling suggests that dust loading on 10–100 km scales can lead to the formation of organized convection powered by the solar heating of dust (Rafkin 2009; Spiga et al. 2013). In the simulations of Spiga et al. (2013), convection in the center of the storm extended significantly above the planetary boundary layer. In simulations by Rafkin (2009), convection was mostly confined to the planetary boundary layer. However, this convection formed a self-sustaining circulation like a tropical cyclone under some conditions (Rafkin 2009). From study of one local dust storm, Kahn (1995) argued that local dust storms require external forcing for growth and maintenance as well as for initiation. This hypothesis has not yet been tested systematically.

Observational studies have mostly focused on synoptic to planetary aspects of dust storm structure, but the mesoscale structure apparent in visible imagery is often described as well (e.g., Briggs et al. 1979; Kahn 1984; Cantor et al. 2001; Strausberg et al. 2005; Cantor 2007). Typically, these descriptions contrast distinct turbulent features such as plumes and cells on scales between the few km resolution of the imagery and ~100 km (“texture”: Guzewich et al. 2015; Kulowski et al. 2016) with indistinct laminar haze. Guzewich et al. (2015) argues that texture indicates circulations associated with active dust lifting, because advection and diffusion would homogenize them into hazes on timescales of hours. Due to the resemblance of some textures to convective clouds on the Earth, texture also has been interpreted as an indicator of deep convection as well as of active lifting (Strausberg et al. 2005). Yet with the possible exception of near-infrared imaging of a single local storm (Määttänen et al. 2009), the connection between dust storm texture and other aspects of mesoscale structure and/or vertical mixing in dust storms has not been investigated. This study therefore systematically investigates the relationship between the appearance of local dust storms in visible imagery and their thermal and aerosol structure in a smooth, low elevation, and dusty region of interest (ROI) in the dusty Martian plains of NE Amazonis and SW Arcadia Planitiae (25°-40° N,150°-165° W).

One challenge of studying dust storm structure on Mars is that most observations of an individual dust storm are limited to a single orbital track, which will intersect only one part of the storm. It is therefore hard to compare and contrast storm structure in regions where the visible presentation of storms strongly varies. In addition, observations by thermal sounders and lidar will be affected by variability in topography, albedo, and thermophysical properties, complicating comparisons of storms of even identical visible appearance over terrain of variable characteristics. Dust storms in the ROI, however, have distinctive visible structure. And the ROI has the additional advantage of being topographically smooth and having albedo and thermophysical properties that vary fairly simply across the region.

2. Methods and Data

a. Martian Time

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 (L_s): the angle between Mars and the Sun relative to the northern vernal equinox (Clancy et al. 2000; Piqueux et al. 2015).

b. Properties of the ROI

The ROI is relatively topographically smooth. There are a few small craters and one unnamed larger crater at 29° N, 153° W (Fig. 1a). Most of the area falls within a 500 m altitude range (Figs. 1a and 2a) typical of Mars’s northern lowlands. The standard deviation of topography in part of this area when mapped at ~ 0.5 km resolution is ~ 20 m (Fenton and Lorenz 2015).

Fig. 1.

Surface properties of the ROI in NE Amazonis–SW Arcadia Planitiae: (a) Topography relative to Mars’s areoid (km) from the Mars Orbiter Laser Altimeter (MOLA) (Smith et al. 2001, 2003). Resolution: 16 points per degree; (b) Reflectivity at 1064 nm based on MOLA observations (Heavens 2016, re-submitted to Icarus) (available from the author by request). Resolution: 2 points per degree; (c) Daytime thermal inertia (J m⁻² s^−1/2) from Thermal Emission Spectrometer (TES) observations (Putzig and Mellon 2007; Putzig et al. 2009). Resolution: 20 points per degree; (d) 1-(Dust Cover Index) from TES observations (Ruff and Christensen 2002; Ruff cited 2016). Resolution: 16 points per degree.

Fig. 2.

Comparisons of the probability distributions of surface properties in the ROI and over 60° S–60° N. Probability is expressed as fraction of total area, either of the ROI or of 0° S–60° N: (a) Topography relative to Mars’s areoid (km) (Smith et al. 2001). Binning interval is 100 m.; (b) Reflectivity at 1064 nm (Heavens 2016, re-submitted to Icarus). Binning interval is 0.01 units; (c) Daytime thermal inertia (J m⁻² s^−1/2) (Putzig and Mellon 2007). Binning interval is 10 J m⁻² s^−1/2; (d) 1-(Dust Cover Index) (Ruff and Christensen 2002). Binning interval is 0.002 units.

Reflectivity, thermal inertia, and DCI (a measure of dust emissivity in the thermal infrared) are proxies for surface dust abundance that are sensitive to different depth ranges (Ruff and Christensen 2002). Visible and near-infrared reflectivity have skin-depths on the order of a few microns (Ruff and Christensen 2002). Moving to thermal infrared wavelengths enables DCI to sample surface dust abundance on the order of tens of microns, while thermal inertia samples on the length scale of the diurnal thermal wave (on the order of centimeters) (Ruff and Christensen 2002).

Away from Mars’s ice caps, surface albedo/reflectivity is bimodally distributed (Fig. 2b). Most of the area would be classified as bright, though a dark reflectivity feature is present on the northern end of the ROI (Fig. 2b). This dark feature follows a boundary between low thermal inertia and intermediate thermal inertia (Fig. 1c), a surface property that is likewise bimodally distributed (albeit with a long tail) (Fig. 2c). The low reflectivity feature also closely maps to a feature of high dust cover index (DCI) (low additive inverse DCI) (Fig. 1d). DCI or its additive inverse are likewise bimodally distributed away from the poles (Fig. 2d). Thus, from least dusty to most dusty are: (1) the eastern part of the low reflectivity feature; (2) the western and central areas of the low reflectivity feature; (3) the higher thermal inertia but still high reflectivity area; and (4) the low thermal inertia and high reflectivity area.

c. Observational Data

This section will briefly describe each type of data. These descriptions will include data processing not covered in the scientific descriptions of the data or in the documentation where the data is archived. Citations to scientific descriptions and data sources are included in an overview in Table 1.

Table 1.

Sources of observational data for this study. Note that all data is subject to a variety of interruptions in temporal coverage or variations in resolution or other aspects of data quality. The cited references under Scientific Description and the documentation accompanying the Data Source should be consulted for further details. Temporal coverage is based on availability at the data source at the time of last access.

Data Type	Spectral Range	Spatial Resolution	Temporal Coverage	Scientific Description	Data Source
MGS-MOC Daily Global Maps	Visible wavelengths	8 points per degree	MY 24, Ls ~150° – MY 28, Ls ~ 122°	Malin and Edgett (2001) Wang and Richardson (2015)	Wang (cited 2016)
MRO-MARCI Daily Global Maps	Visible wavelengths	8 points per degree	MY 28, Ls ~ 132° – MY 30, Ls ~ 111.8°	Bell et al. (2009) Wang and Richardson (2015)	Wang (cited 2016)
MRO-MARCI Raw Imagery	UV–NIR wavelengths	0.7–4 km	MY 28, Ls ~ 110° – MY 33, Ls ~ 75°	Bell et al. (2009)	MARCI (cited 2016)
MGS-TES Radiance	200 cm−1–1600 cm−1 5–10 cm−1 resolution	~ 3 km	MY 24, Ls ~ 103° – MY 28, Ls ~ 121°	Christensen et al. (2001)	Christensen (2002)
MGS-TES Retrievals	Temperature: ~ 15 μm Dust: 1075 cm−1 Water ice: 825 cm−1	along-track: 10–20 km; cross-track: 9 km	MY 24, Ls ~ 103° – MY 28, Ls ~ 121°	Conrath et al. (2000) Smith (2004)	Christensen (2002)
MGS-MOLA Active Radiometry	1064 nm ± 2 nm	spot size: 150 m; horiz. spacing: 300 m	MY 24, Ls ~ 104° – MY 25, Ls ~ 186°	Smith et al. (2001) Neumann et al. (2003)	MOLA (cited 2016a)
MGS-MOLA Passive Radiometry	1064 nm ± 2 nm	during active sounding: 0.34 km × 3 km; after active sounding: 0.34 km × 0.5 km (cross-track × along-track)	MY 24, Ls ~ 104° – MY 28, Ls ~ 116°	Sun et al. (2006)	MOLA (cited 2016b)
MRO-MCS Limb Radiance	Broadband channels Visible–TIR	~ 5 km vertical ~ 6 km horiz. spacing	MY 28, Ls ~ 111° – MY 32, Ls ~ 334°	McCleese et al. (2007)	MCS (cited 2016b)
MRO-MCS Retrievals	Temperature: ~ 15 μm Dust: 463 cm−1 Water ice: 843 cm−1	~ 5 km vertical ~ 33 km horiz. spacing	MY 28, Ls ~ 111° – MY 32, Ls ~ 334°	Kleinböhl et al. (2009, 2015)	MCS (cited 2016a)

Most of the visible imagery is calibrated, photometrically-corrected and equirectangular projected Daily Global Maps from observations by the Mars Orbiter Camera (MOC) on board Mars Global Surveyor (MGS) and the Mars Color Imager (MARCI) on board Mars Reconnaissance Orbiter (MRO). However, MARCI images can be higher resolution than 8 points per degree (Table 1). In addition, MOC and MARCI must be processed differently because of camera differences (Wang and Richardson 2015). Where resolution or processing differences were of interest, raw MARCI imagery was obtained. Using the United States Geological Survey’s (USGS) Integrated Software for Imagers and Spectrometers (ISIS) package, images were calibrated, photometrically corrected with a Minnaert function with a k parameter of 0.7 to emphasize the presence of dust (Cantor 2007), and projected in equirectangular coordinates. Raw data with emission or solar incidence angles greater than 80° were excluded from the projected image.

To complement MOC imagery, calibrated radiances (converted to brightness temperature), retrieved temperature profiles from the surface to 40 km, and absorption-only column opacity retrievals from nadir observations by the Thermal Emission Spectrometer (TES) on board MGS were used. Radiance was measured by three pairs of two detectors that formed observational tracks that are ~ 3 km apart on the surface. A surface temperature was retrieved from each spectrum by the TES team (Christensen et al. 2001). The retrieved temperature profiles and column opacity retrievals were based on averages of data from measurements by all 6 detectors. TES nadir retrievals were not made over areas with poor thermal contrast between the surface and the atmosphere, which limited retrievals at night and over the winter pole.

Observations by MOLA on board MGS were used to complement MOC imagery. MOLA observed in an active radiometry mode, where it measured the time-of-flight and returned power of a laser pulse (“shot”) emitted by the instrument at nadir toward the surface. The product of the reflectivity and two-way atmospheric transmissivity ($r_s{t}_{atm}^2$, also written RT) can be derived from each measurement using the lidar link equation (Neumann et al. 2003). For surface returns where the surface reflectivity (r_s) is known, the column opacity at 1064 nm (τ₁₀₆₄) is:

$${\tau }_{1064}=-\frac{1}{2}\log (\frac{RT}{r_s})$$ (1)

MOLA cannot differentiate between dust and ice. When independent information suggests that dust is the dominant source of opacity, MOLA opacity can be converted to TES equivalent by dividing by a factor of 2.6 with an uncertainty of up to 30% (Montabone et al. 2015; Heavens 2016, re-submitted to Icarus). The analogous conversion for water ice can be accomplished by dividing by ~ 3 (Clancy et al. 2003a; Smith 2004).

Reflectivity at 1064 nm was estimated from the 0.5° × 0.5° map of Heavens (2016, re-submitted to Icarus) (Heavens cited 2016). Under clear conditions over most surfaces, laser pulse returns exceeded the digital range of the MOLA detector, which resulted in RT being a minimum estimate. For saturated returns, opacity estimated from Eq. 1 is an upper bound, while the lower bound is zero. As the strength of MOLA’s laser diminished, saturation was less common (Neumann et al. 2003). The uncertainty in saturated and unsaturated returns is fully discussed by Heavens (2016, re-submitted to Icarus), but note that the 1σ uncertainty in column opacity is ≤ ± 0.1 for unsaturated surface returns. As a result of uncertainty about aerosol forward scattering to the MOLA detector, it is possible that Eq. 1 may underestimate column opacity over the ROI during dust storms Heavens (2016, re-submitted to Icarus). The impact of this bias on this study would lead to underestimate of the column opacity variations described in Section 3.

MOLA’s passive radiometry mode measured Mars’s brightness (presumably reflected sunlight) at 1064 nm. This data was converted to a Minnaert model-based normal albedo (k=0.7) and re-calibrated, so that it could be compared with the 1064 nm reflectivity map of Heavens (2016, re-submitted to Icarus). The apparent brightness of Mars at 1064 nm is relatively insensitive to opacity at low opacity but quite sensitive at opacities > 1 (Clancy et al. 2003b; Szwast et al. 2006; Heavens 2016, re-submitted to Icarus), so passive radiometry may be a useful indicator of relative opacity variations in dust storms during the day.

Limb radiance observations and retrievals of vertical temperature, dust opacity, and water ice opacity profiles (from the surface to 80 km) from nadir/off-nadir and limb observations by MCS on board MRO were used to complement MARCI imagery. Vertical profiles are retrieved on pressure coordinates, which can be referenced to an altitude grid relative to the surface and derived from the pointing of the instrument. The not yet publicly available version 5 retrievals (Kleinböhl et al. 2016) are presented in this study, but it has been verified that using Version 4 (listed in Table 1) does not materially change the results.

Retrieved opacities were converted to density-scaled opacity, a quantity proportional to mass mixing ratio within uncertainties about aerosol properties. The density-scaled opacity in m² kg⁻¹ is 12000 × the dust mass mixing ratio in ppm (Heavens et al. 2011a, 2015). The water ice opacity must be divided by a factor of 1.5 when comparing with TES to account for the difference between absorption-only and extinction opacities (wavelengths are similar) (Smith 2004). MCS dust opacity must be multiplied by a factor of 2.7 to be compared with TES (Montabone et al. 2015), 7.3 to compare with a visible opacity at 660 nm (Heavens et al. 2011a), and therefore ~ 7.8 to compare with opacity at 1064 nm (following Ockert-Bell et al. (1997) and Clancy et al. (2003b)).

d. Climatological Guidance in Support of Interpretation

Data from the Mars Climate Database (MCD) v5.2 (a reference atmospheric database based on GCM output) was downloaded for two scenarios: climatology average solar scenario (CASS) and dust storm average solar scenario (DSASS) (Forget et al. 1999; Millour et al. 2015; MCD cited 2016). The CASS data is appropriate guidance for the circulation and boundary layer properties when unperturbed by the local dust storm. The DSASS assumes a globally uniform visible dust opacity of 5, and so is appropriate for estimating a dust storm’s effects on convective boundary layer height (h).

Hourly binned meteorological data from the Viking Lander 2 (VL2) during MYs 12 and 13 provided guidance about surface wind speed and direction (Hess et al. 1977; Tillman and Johnson 1997). VL2 observed near 48° N, 134° E, a site well outside the ROI but in a similarly low elevation area in the northern mid-latitudes. Thus, this is the dataset of surface wind observations most easily compared with the ROI.

e. Focus of the Study/Dust Storm Survey of the ROI

This study focuses on textured local dust storms within the ROI (or mostly so) during the early-mid afternoon (13:00-16:00 Local Solar Time (LST)), as determined by survey of MOC and MARCI imagery during two Mars Years, one during the MGS era (MY 24), and one during the MRO era (MY 29). Textured local dust storms that appear directly associated with frontal boundaries were excluded, because these storms produce greater dust cloud cover outside the ROI than inside it. This focus optimizes the observing of textured local dust storms to periods from which the fullest complement (as measured by number and functionality of individual instruments) of spacecraft observations were available and global dust storm activity was absent.

Note that observations from MGS instruments did not begin until early northern summer of MY 24. The potential for textured local dust storm activity in the ROI prior to the commencement of observations by MGS will be considered in the next section.

3. Results

a. Climatology

All but one of the textured local dust storms in the ROI occur during two periods: L_s = 190°–230° (early northern fall) and L_s = 290°–340° (the middle of northern winter) (Fig. 3). The first period is far less active than the second period in MY 24, but both periods are similarly active in MY 29. Activity is minimal ~ L_s = 270° (northern winter solstice).

Fig. 3.

Climatology of textured local dust storms in the ROI during MY 24 and MY 29. The dates and data availability of storms are indicated with colored crosses. Sounder data indicates that the storm was observed by TES, MCS, and/or MOLA. The start or end of observations by MOC or MARCI during the relevant year are indicated with colored dashed lines.

The exception is the unusually early storm observed at L_s = 147.53° in MY 29 (Fig. 3). During MY 29, local and regional dust storm activity (normally minimal in northern spring and summer) commenced earlier than normal (~ L_s = 135°) (Smith 2009; McCleese et al. 2010; Heavens et al. 2011a). Such “early season dust storm activity” was not observed during MY 24 (Smith 2009) but would have been observable in TES observations (from L_s = 104°) before MOC daily global mapping commenced. In fact, at least two storms were observed just beyond the eastern margin of the ROI in the L_s = 140°–150° period (Cantor et al. 2006). However, the L_s = 147.53° in MY 29 storm is well within the ROI. The potential link between early season dust storm activity and local dust storm seasonality in the ROI suggests that it is unlikely that storms occurred in MY 24 prior to MOC daily global mapping. Several storms occurred in MY 24 after L_s = 327°, so it is possible that a few storms were missed during the interruption in MARCI coverage in MY 29 (Fig. 3). However, the fundamental climatology is unaltered by the gaps in coverage.

b. Visible Presentation of Storms

Storms vary in both morphology (shape and extent of the storm as a whole) and texture (cloud top appearance in terms of structure, lack of structure, and contrast with the surrounding surface). Storms can be up to 1500 km long and 500 km wide (Fig. 4a), an isolated cloud no more than 100 km in any dimension (Fig. 4b), or quasi-rectangular and of intermediate size (typical) (Fig. 4c). The northern boundary of most storms is the dark reflectivity feature in the ROI (Figs. 4a-c,e). Some storms appear to curve around this feature (Figs. 4a and c), while others are aligned along a straight line parallel to it (Fig. 4b). However, one storm expanded to the north of it (Fig. 4d) and a few storms are observed far to the south of it (Fig. 4f).

Fig. 4.

Examples of textured dust storms in the ROI (season of observation indicated by the panel titles) during MYs 24 and 29.

Storms also have a variety of textures. The storm in Fig. 4d resembles the cumuliform dust clouds previously reported at Mars (Strausberg et al. 2005): it has a puffy or smoky texture with some embedded clouds of ~ 100 km diameter. However, the most distinctive texture of storms in the ROI consists of linear features or bands that are approximately aligned in parallel. This ruffled texture can be seen in the northeast of the storm in Fig. 4a.

The clearest examples of ruffled texture are from three storms during MY 24 (Figs. 5a-f). The MOC imagery resolves alternating bright and shadowed bands with some curvature or bifurcation. The wavelength of this banding is ~ 30 km (4× the resolution of the image). In Fig. 5d, the direction of banding is southwest–northeast, while it is approximately south–north in Figs. 5e-f. Some storms may have ruffled texture that is only partially resolved at the resolution of the imagery, as could explain the appearance of two small areas of alternating bright and shadowed oval clouds in Fig. 4c). Other storms are too small for texture to be assessed (Fig. 4b).

Fig. 5.

Examples of dust storms with ruffled texture from MY 24 (season of observation indicated by the panel titles). The top row contains an image of the entire storm and a box outlining the bounds of the inset displayed in the panel directly below.

In the MARCI imagery of Wang and Richardson (2015), no storms have a ruffled texture, but several have a honeycomb texture (made up of rectangular cells with bright edges) not seen in the MY 24 storms imaged by MOC. At higher resolution and with a simpler photometric correction, one example of honeycomb texture (Fig. 6a) becomes ruffled texture with wavelengths up to 60 km and some embedded oval clouds (40 × 20 km) (Fig. 6d). So poorly resolved ruffled texture may explain the honeycomb appearance of other storms. In addition, two storms with non-honeycomb texture (Figs. 6b-c) also appear ruffled (east-west orientation) at higher resolution (Figs. 6e-f). Ruffled texture is dominant in 12 of 29 (41%) of storms in the ROI (Table 2).

Fig. 6.

Examples of dust storms with ruffled texture from MY 29 (season of observation indicated by the panel titles). The top row contains an image of the entire storm in imagery processed by Wang and Richardson (2015) (8 points per degree) and a box outlining the bounds of the inset displayed in the panel directly below. The bottom row contains MARCi imagery processed as described in Section 2 at a resolution of 12 points per degree. The lower panels have been stretched to enhance the contrast between light and dark areas and the contrast of all panels has been uniformly increased.

Table 2.

List of storms found by the survey defined in Section 2e. The L_s in which the storm is observed and the mean L_s of the daily global map (as given in the databases associated with Wang and Richardson (2015) are both listed. The assessment of ruffled texture counts only the unambiguous cases at the best available resolution.)

MY	Ls (storm)	Ls (daily global map)	Sounder Data?	Ruffled?
24	210.11	210.15	Y	Yes
24	219.57	219.22	N	Yes
24	312.53	312.22	N	Too small to tell
24	314.89	314.65	N	Yes
24	317.85	317.65	Y	Yes
24	323.67	323.43	N	Too small to tell
24	324.23	323.98	Y	Yes
24	331.05	331.04	N	No
24	332.75	332.65	N	No
24	333.33	333.08	N	Yes
24	334.45	334.25	Y	Too small to tell
24	335.54	335.17	Y	No
24	337.75	337.51	Y	Yes
29	147.53	147.4	Y	No
29	199.02	199.2	Y	Yes
29	203.30	203.2	N	Yes
29	209.51	209.4	Y	No
29	210.79	210.6	Y	No
29	212.64	212.8	Y	No
29	225.36	225.5	Y	No
29	227.95	227.8	Y	No
29	291.80	291.9	Y	No
29	300.50	300.3	Y	No
29	302.31	302.5	Y	Yes
29	304.76	304.8	Y	No
29	312.56	312.4	N	No
29	320.81	320.8	Y	Yes
29	322.57	322.8	N	No
29	327.15	327.0	Y	Yes

c. Horizontal Structure

Storms produce noticeable drops in surface temperatures underneath their clouds, the consequence of strongly reduced surface insolation in an atmosphere with a radiative relaxation timescale 40–60× smaller than the Earth’s (Goody and Belton 1967; Read et al. 2015) and the low thermal inertia (Fig. 1c). There are typically breaks in TES observations/retrievals of storms (Figs. 7a-d) and little data poleward of 45° N, but the extant data associates the storms with anomalous temperature minima of 10–30 K relative to a general poleward decrease in temperature (Fig. 7e). The smallest drops are associated with the observations on storm margins (Figs. 7c and e). Similar drops are observed by MCS in the MY 29 storms (Fig. 8). There is a 30 K drop in the unusually low-latitude storm in Fig. 8c (Fig. 8e), while the unusually early storm is associated with a weak poleward surface temperature gradient (Figs. 7e and 8e).

Fig. 7.

Surface temperature retrievals from TES near and over dust storms in the ROI in the context of MOC imagery. (a)-(d): Context images with the positions of TES retrievals indicated by the markers; (e) Surface temperature vs. latitude, as labeled.

Fig. 8.

Surface temperature retrievals from MCS near and over dust storms in the ROI in the context of MOC imagery. (a)-(d): Context images with the positions of MCS retrievals indicated by the markers; (e) Surface temperature vs. latitude, as labeled.

The storms also are associated with high TES infrared dust column opacities (Figs. 9e-h). Dust opacity generally exceeds 1 in storm centers (Figs. 9e,f,and h) and is greater than 0.5 on storm margins (Figs. 9g). In some cases, dust storms are associated with a maximum in water ice opacity (Figs. 9e-f). In other cases, water ice opacity changes little in the dust storm (Figs. 9g-h). However, in all cases, dust is the dominant source of 1064 nm opacity inside the storm and is likely to be the dominant source of 1064 nm opacity outside of it.

Fig. 9.

Horizontal opacity structure of example storms in MY 24: (a)-(d): Context images with the positions of TES retrievals indicated by the black markers; (e-h) Retrieved TES dust and water ice column opacities in the vicinity of the dust storm shown in the panel directly above.

These high opacities (equivalent to > 2.6 at 1064 nm) limit the utility of MOLA active sounding measurements in storm centers. Typical surface reflectivities in the ROI ≈ 0.4 (Fig. 2b). As a result of absorption of the MOLA laser, it is difficult to measure RT values < 0.02 (Neumann et al. 2003). So 1064 nm opacities > 1.5 will be difficult to measure.

Available MOLA observations oriented approximately perpendicular to ruffled texture (e.g., the storms in Figs. 9a and b) show that the ruffled texture is associated with fine structure in column opacity. In one storm (Fig. 10), hints of a secondary peak in dust opacity on the southern margin of the storm are observed in TES retrievals (Fig. 9f). However, MOLA active sounding resolves four (and part of another) oscillations (with amplitude up to 0.7) in column opacity on the southern margin of the storm (Fig. 10b) and two oscillations in opacity on the northern margin of the storm (Figs. 10b-c). The minima of similar oscillations in opacity may be visible at 37.3° N, 37.7 ° N, and 38.0 ° N (Figs. 10b-c). The first set of oscillations corresponds to a region where the ruffled texture is hard to distinguish in visible imagery (Fig. 10a). The wavelengths of the oscillations on the southern margin of the storm shorten to the north and are even smaller on the northern margin of the storm. Passive radiometry also suggests that this part of the storm (31.5 ° N–34 ° N) is similarly bright to the surface. Yet there are small variations in brightness that correlate with the peaks in opacity (Figs. 10b and d). As conditions become too opaque for active sounding (northward of 34 ° N), oscillations in reflectivity in passive radiometry continue and the reflectivity increases (Fig. 10d), in line with the ruffled texture and increased brightness of the visible image (Fig. 10d). The extended dimness of the area north of the storm relative to the surface may indicate the shadow of the storm or uncertainties in the surface reflectivity map.

Fig. 10.

(a) MOC image of a portion of a ruffled storm (MY 24, L_s = 317.85°) with the location of MOLA returns indicated with black crosses; (b) 1064 nm aerosol column opacity in the vicinity of the storm estimated from MOLA active sounding data. Missing data is a result of non-surface returns, from which a column opacity cannot be estimated. Black dots indicated saturated returns, while red crosses indicate unsaturated returns; (c) 1064 nm aerosol column opacity, as in (b), but focused on the northern margin of the storm; (d) Normal reflectivity in the vicinity of the storm estimated from MOLA passive radiometry data. The black line indicates the estimated surface reflectivity at 0.5° resolution (Heavens 2016, re-submitted to Icarus).

The wavelengths of the oscillations (and thus the scale of the ruffled texture) were quantified by measurement of their peak to trough distance in latitude (corrected by ~ 5° orientation of the ruffles off of east–west) and doubling. The first four oscillations on the southern margin have wavelengths of 70 km, 42 km, 32 km, and 24 km respectively. The two resolved oscillations on the northern margin have wavelengths of 10 km and 9 km.

In another storm (Fig. 11), MOLA observations are on the northeastern side of the storm, just on the margin of dust clouds with ruffled texture (Fig. 10). MOLA active sounding observes three (or perhaps up to five) oscillations in column opacity as well as another dusty area outside of the ruffled texture at 42° N (Figs. 11b and c). Passive radiometry indicates that the area is dimmer than normal to the northeast of the storm (a possible shadowing effect), but there is a brightness oscillation associated with the dusty area at 42° N (Fig. 11d).

Fig. 11.

(a) MOC image of a portion of a ruffled storm (MY 24, L_s = 324.23°) with the location of MOLA returns indicated with black crosses; (b) 1064 nm aerosol column opacity in the vicinity of the storm estimated from MOLA active sounding data. Missing data is a result of non-surface returns, from which a column opacity cannot be estimated. Black dots indicated saturated returns, while red crosses indicate unsaturated returns; (c) 1064 nm aerosol column opacity, as in (b), but focused on the region of closest proximity between the storm and MOLA observations; (d) Normal reflectivity in the vicinity of the storm estimated from MOLA passive radiometry data. The black line indicates the estimated surface reflectivity at 0.5° resolution (Heavens 2016, re-submitted to Icarus).

TES observed this storm to the west of MOLA, yielding the only continuous TES radiance observations perpendicular to ruffled texture (Fig. 9c). TES surface temperatures vary the most where the oscillations in column opacity are observed (Figs. 12b and d). Accounting for detector arrangement, four troughs in surface temperature are resolved (Fig. 12c). These troughs are deeper, the farther west the observations are made. The northernmost three of the troughs are at the approximate latitude of the column opacity peaks (Figs. 12b and d). Therefore, minima in surface temperature closely align with maxima in column opacity. The approximate wavelength of the ruffled texture here is 24 km.

Fig. 12.

(a) MOC image of a portion of a ruffled storm (MY 24, L_s = 324.23°) with the location of TES radiance and MOLA observations, as indicated by the legend. The notation “x+y” indicates detectors x and y; (b) Surface brightness temperature (K) retrieved from TES radiance observations in the vicinity of the storm. Marker colors are as in the legend; (c) 1064 nm aerosol column opacity in the vicinity of the storm estimated from MOLA active sounding data. Saturated returns are the low opacity data. (d) Normal reflectivity in the vicinity of the storm estimated from MOLA passive radiometry data. The black line indicates the estimated surface reflectivity at 0.5° resolution (Heavens 2016, re-submitted to Icarus).

d. Vertical Structure

The vertical extent of the storms can be inferred from their impact on the thermal structure and vertical dust distribution of the atmosphere. While the dust storms studied here have a significant impact on surface temperature, they only impact atmospheric thermal structure within the first atmospheric scale height. TES retrievals at the highest pressure level typically reported in this area (and therefore nearest the surface) only have a negative poleward slope (Figs. 13a-d). There is no evidence for atmospheric heating by dust, nor for the temperature depression seen in the surface temperature retrieval (Figs. 13a-d).

Fig. 13.

Temperature variability near four example dust storms. Each panel shows TES retrievals of temperature (K) at the surface and 783 Pa (red and black markers) as well as the estimated surface pressure (Pa) (blue line).

The limited vertical extent of the storms is also implied by TES radiance data. Consistent with the surface temperature retrievals, areas outside a storm or within less dusty areas inside a storm have uniformly higher brightness temperatures in continuum emission outside the 15 μm (667 cm⁻¹) band of CO₂ (Fig. 14). In addition, less dusty areas have uniformly higher brightness temperatures inside the 15 μm band of CO₂, though brightness temperatures in less dusty areas are < 5 K warmer within 30–40 cm⁻¹ of the band center (Figs. 14b and d). First, this contrast in spectra suggests that the dust storm has an overall negative effect on atmospheric temperature. Second, the much smaller difference between dust storm and non-dust storm spectra in the 15 μm band of CO₂ suggests that this negative effect is fairly shallow. Contribution functions for atmospheric temperature in the 15 μm band of CO₂ peak at 500 Pa at ~ 40 cm⁻¹ from the band center and at 300 Pa at ~ 30 cm⁻¹ from the band center (Conrath et al. 2000). Therefore, most cooling is at pressures higher than between 300–500 Pa. Estimated surface pressures in the vicinity of these dust storms are ~ 800 Pa (Fig. 14). Therefore, the impact of the storm on the thermal structure is limited to the first atmospheric scale height or less. The TES atmospheric temperature retrievals may not capture the entirety of this cooling, because they are smoothed at a resolution of 0.75 scale heights (Conrath et al. 2000).

Fig. 14.

Differences between two TES spectra (converted to brightness temperature (K)) within and near two example dust storms. The blue and red circles in the left panels indicate the location of the spectra in MOC imagery. The distinction between inside the storm and outside the storm in the top panel and between more or less dusty conditions in the bottom panel is based on surface temperature. The right panels show the difference between the spectra at the red and blue points in the respective left panel. The vertical dotted lines indicate ±30cm⁻¹ and ±40cm⁻¹ from the center of the 15μm CO₂ band.

The impact of storms on thermal structure also can be characterized using MCS data by estimating a background temperature field from interpolation to the north and south of a dust storm (the red points in Fig. 15a) and differencing it from a temperature retrieval near the dust storm center (the blue point in Fig. 15a). This storm cools the surface by ~ 30 K (Fig. 15c) but has a cooling impact of 1–2 K in the first scale height above the surface (Fig. 15c). The storm has no apparent impact on the local temperature field (Fig. 15b), so the larger temperature difference at pressures lower than 100 Pa are probably due to atmospheric variability unrelated to the dust storm. While it is possible that the dust storm cooling is poorly resolved because of the broad weighting functions associated with MCS retrievals at low altitudes (Kleinböhl et al. 2009), the temperature impact of the dust storm still appears to be restricted to the first atmospheric scale height.

Fig. 15.

Estimated effect of a local dust storm on a MCS temperature profile: (a) Surface temperature data near the surface (black crosses). The red circled data (and the higher altitude data at the same locations) are used to estimate the background temperature structure by linear interpolation. The blue circled data (and the higher altitude data at the same location) is taken to be the dust storm temperature structure; (b) MCS retrieved temperature (K) structure in the vicinity of the dust storm; (c) Difference between the dust storm temperature profile and the background temperature profile (K).

Dust in most storms is confined to the first atmospheric scale height as well. A typical example is shown in Fig. 16. Far to the south of the storm, there is a DDL with mass mixing ratio of up to 30 ppm, but mass mixing ratios over the storm are less than 10 ppm (Fig. 16). Retrievals over the storm are as low as 8 km above the surface, and the total 1064 nm column opacity inferred from the MCS dust retrievals is ~ 0.35, much less than the > 1.5 or > 2.6 implied by MOLA and TES observations of the centers of dust storms in the ROI during MY 24. Thus, a 1064 nm column opacity of ~ 2 must be confined below 8 km. A well-mixed dust distribution to some height z_c has an opacity of approximately:

$$\tau ={\rho }_0{\left(\frac{d{\tau }_z}{\rho }\right)}_0\underset{0}{\overset{z_c}{\int }}\text{exp}\left(\frac{-z}{H}\right)$$ (2)

where H is the atmospheric scale height, ${\left(\frac{d{\tau }_z}{\rho }\right)}_0$ is the surface dust density-scaled opacity, and ρ₀ is the surface air density. Note that the dimensions of opacity (dτ_z) are inverse length. Assuming a 1064 nm column opacity of 2, surface pressure of 800 Pa, a surface temperature of 220 K, and an atmospheric scale height of 10 km, the expected dust mass mixing ratio in the storm is 29 ppm. That such a high dust mass mixing ratio is not observed demonstrates that the dust in the storm is confined below 8 km. Note that if the particle size of dust were larger in the storm than assumed, the ratio of 1064 nm to MCS dust opacity would decrease (Clancy et al. 2003a; Kleinböhl et al. 2009), resulting in lower inferred 1064 nm column opacity and higher expected dust mass mixing ratio than calculated here.

Fig. 16.

Dust distribution in the vicinity of the labeled dust storm: (a) Dust distribution in the vicinity of the dust storm expressed as mass mixing ratio (ppm) estimated from MCS retrievals. The location of the storm center indicated by surface temperature data is marked with a black circle. Areas where data is interpolated beyond the normal spacing of MCS retrievals have been covered by whitespace; (b) The lowest altitude above the surface at which data is reported for MCS dust retrievals in the vicinity of the dust storm. The datum for the center of the storm is marked with a black circle; (c) integrated opacity of the dust retrievals in the vicinity of the dust storm converted to 1064 nm opacity. The datum for the center of the storm is marked with a black circle.

A few storms may extend to higher altitudes. A dust mass mixing ratio of ~ 25 ppm is observed 20 km above the surface on the southern margin of one storm (Figs. 17a-b). Mass mixing ratios of up to 100 ppm are observed 40 km above the surface in the unusually early dust storm, while a DDL of 30–40 ppm is observed to the north of this storm (Figs. 18a-b). Visible imagery and the failure of the surface temperature retrieval suggests total column opacity is high (Fig. 8a and e). Yet this storm may not have been a vertically continuous dust cloud; the failure of MCS retrievals at 40 km and below could be due to an unusually thick DDL confined around 40 km lying over a storm confined to the boundary layer. In addition, the absence of retrieval information below 40 km does not allow the storm’s temperature impact to be determined.

Fig. 17.

Dust distribution in the vicinity of the labeled dust storm: (a) Dust distribution in the vicinity of the dust storm expressed as mass mixing ratio (ppm) estimated from MCS retrievals. The location of the storm center indicated by surface temperature data is marked with a black circle. Areas where data is interpolated beyond the normal spacing of MCS retrievals have been covered by whitespace; (b) The lowest altitude above the surface at which data is reported for MCS dust retrievals in the vicinity of the dust storm. The datum for the center of the storm is marked with a black circle; (c) integrated opacity of the dust retrievals in the vicinity of the dust storm converted to 1064 nm opacity. The datum for the center of the storm is marked with a black circle.

Fig. 18.

Dust distribution in the vicinity of the labeled dust storm: (a) Dust distribution in the vicinity of the dust storm expressed as mass mixing ratio (ppm) estimated from MCS retrievals. The location of the storm (as indicated by comparison with visible imagery and surface temperature retrieval failure) is marked with black circles. Areas where data is interpolated beyond the normal spacing of MCS retrievals have been covered by whitespace; (b) The lowest altitude above the surface at which data is reported for MCS dust retrievals in the vicinity of the dust storm. The data for the storm are marked with black circles; (c) Integrated opacity of the dust retrievals in the vicinity of the dust storm converted to 1064 nm opacity. The data for the storm are marked with black circles.

In another case, a DDL is observed over a dust storm, but the bulk of the dust storm’s dust still appears to be confined below 8 km (Figs. 19a-c). Directly attributing the DDL to the storm is not straightforward. Figs. 16a and 17b show that DDLs are common on the southern margin of the ROI. Observations on the sol prior to the dust storm show a DDL in the ROI at 25° N (Fig. 20a) of similar magnitude to that observed over the storm (Fig. 20b), yet the layer on the prior sol does not appear near a textured dust storm. Caution thus must be exercised in using DDLs as the sole indicator of dust storm vertical mixing. The most that can be said in favor of deeper vertical mixing is that the storm was mostly confined to the first atmospheric scale height but a few powerful updrafts may have existed within it.

Fig. 19.

Dust distribution in the vicinity of the labeled dust storm: (a) Dust distribution in the vicinity of the dust storm expressed as mass mixing ratio (ppm) estimated from MCS retrievals. The location of the storm center indicated by surface temperature data is marked with a black circle. Areas where data is interpolated beyond the normal spacing of MCS retrievals have been covered by whitespace; (b) The lowest altitude above the surface at which data is reported for MCS dust retrievals in the vicinity of the dust storm. The datum for the center of the storm is marked with a black circle; (c) integrated opacity of the dust retrievals in the vicinity of the dust storm converted to 1064 nm opacity. The datum for the center of the storm is marked with a black circle.

Fig. 20.

Detached dust layering on the sol prior to and the sol of the L_s = 327.15 dust storm of MY 29: (a) MOC imagery in the vicinity of the dust storm on the previous sol on which has been plotted the magnitude (ppm) and approximate geopotential/areopotential height (km) of the peak mass mixing ratio estimated from MCS retrievals (colored dots). The color of the dot indicates the height and the size indicates the magnitude; (b) as in (a) but for the sol of the storm.

However, comparison of the limb radiance data around both storms (Figs. 21a-b) suggests that the unusually early storm was associated with a continuous, deeply mixed layer of dust rather than a narrow DDL. Radiance observations in the dust-sensitive A5 channel normally decrease steeply and uniformly with height, because the atmosphere emits weakly in A5 unless aerosol is present. At or below the surface, A5 observes surface emission. As aerosol opacity decreases with height, so does emission from aerosol in A5. Some additional radiance may come from scattering into the limb or from the surface into the field-of-view wings of the detector (Kleinböhl et al. 2009, 2011, 2015). And thus, A5 radiance should follow latitudinal surface temperature trends. DDLs change this simple picture by reducing or even reversing the decay of opacity with height, resulting in additional emission that can be seen as a weaker (or even reversed) lapse rate in brightness temperature or a horizontally contrasting feature (Heavens et al. 2015). Such a feature can be seen at 30–40 km above the surface in Fig. 21b between 27° and 37° N, the location of the DDL resolved in Fig. 19a. The unusually early storm (centered at 35° N is associated an unusually weak gradient of brightness temperature above the surface that horizontally contrasts with the surrounding atmosphere sufficiently well to form a bump. Detailed interpretation of this feature would require a radiative transfer model (Fig. 21a), but the considerable emission between the surface and 35 km implies dust opacity is high throughout this range, not just near the surface and 40 km. Radiance data in temperature-sensitive channels (omitted for brevity) do not show a significant atmospheric temperature effect.

Fig. 21.

Observed A5 channel (centered at 463 cm⁻¹) brightness temperature (K) observed by MCS in the vicinity of two dust storms, as labeled. The equivalent retrieved dust fields are in Figs. 18 and 19.

4. Discussion

In this section, it is proposed that most of the dust storms in the ROI are structurally and dynamically analogous to wide mixed layer rolls in the Earth’s atmosphere. Alternative hypotheses and exceptional cases also will be considered. The applicability of this result to the rest of Mars then will be discussed.

a. Ruffled Texture: Mediated by Rolls or Gravity Waves?

Ruffled texture resembles cloud streets in the Earth’s atmosphere, which are “quasi-two-dimensional [structures] (i.e., nearly linear)”, “exhibit coherent up- and downdrafts”, and “are a common feature in the atmospheric boundary layer (ABL) under windy conditions” (Young et al. 2002). Implied in “quasi-two-dimensional” is that they are periodic. Cloud streets arise from the alignment of three-dimensional convection cells toward the direction of the mean wind shear, whereby they become two-dimensional convective rolls (Young et al. 2002).

The ruffled texture certainly appears two-dimensional. There is no way to determine whether the ruffles are regions of alternating vertical wind direction. However, if dust lifting rates do not spatially vary, areas of convergent upward flow will be dustier than areas of divergent downward flow. Thus, the opacity variations in these storms are consistent with the hypothetical vertical wind structure. Moreover, most of these storms are confined to the first atmospheric scale height, strongly suggesting that they are a boundary layer phenomenon. Indeed, the dust storm with the greatest vertical extent (the unusually early storm of L_s = 147.53, MY 29) is unambiguously non-ruffled (Fig. 4d).

Yet not all quasi-two-dimensional structures are cloud streets. Kahn (1984) identified two other types on Mars: lee waves and wave clouds. Lee waves, in the sense of Kahn (1984), are one-dimensional periodic clouds generated by gravity wave trains associated with obstacles such as craters or volcanoes. Wave clouds resemble lee waves but are not associated with obstacles. They thus also result from gravity wave trains, just ones excited by sources other than topography (e.g., Fritts and Alexander 2003). The distinction between wave clouds and cloud streets made by Kahn (1984) was that the latter have “double periodicity”, likely referring to the “string of pearls” appearance of some cloud streets when individual cumuli are organized into lines (Young et al. 2002). Confusingly, recent work by Kulowski et al. (2016) classifies textured dust storm activity in terms of “pebbled”, “puffy”, and “plume-like.” The storm in Figs. 5a and d is cited as an example plume-like storm and the ROI is mapped as a hotspot of plume-like storm activity.

Nevertheless, ruffled texture is a novel Martian cloud type that does not fit into the Kahn (1984) scheme and may be improperly classified by Kulowski et al. (2016). Periodicity distinguishes ruffled texture from the elongated, sharp-edged plume clouds and elongated, smooth-edged streak clouds previously reported in dust storms by Kahn (1984). Lee wave clouds are common on Mars (Pickersgill and Hunt 1982; Kahn 1984). On one hand, they have wavelengths of 30–60 km, like ruffled texture. On the other hand, they are only known as condensate clouds (Kahn 1984). Moreover, their dynamics cannot explain the ruffled texture. There is no topography of sufficient size, amplitude, and direction in the ROI for the ruffled texture to be an obstacle to form them (Fig. 1a). The ruffled texture more strongly resembles wave clouds and cloud streets, which, too, were only observed in condensate clouds by Kahn (1984).

Wave clouds and cloud streets can be difficult to disambiguate. The convective rolls that make up cloud streets can interact with and even generate gravity waves above the cloud layer (Balaji et al. 1993; Young et al. 2002; Magalhaes et al. 2011; Melfi and Palm 2012). Moreover, mismatch between the length scales of gravity waves and boundary layer convection can result in suppression of convection in the wave troughs (Balaji and Clark 1988). Thus, the propagation of gravity wave trains through even disorganized boundary layer convection might look like convective rolls. One difference between these phenomena is how they interact with the wind. The long axis of a convective roll aligns with the boundary layer wind shear vector, while gravity wave trains would propagate along the mean wind in the inversion above the boundary layer. Thus, if rolls and waves were simultaneously present, their alignment would require strong rotational shear perpendicular to the long axis direction of the ruffled texture (Young et al. 2002).

Gravity waves are an imperfect explanation for the length scales of ruffled texture. On one hand, the wavelength change across the storm in Fig. 10 could be due to resonant interaction between gravity waves and boundary layer convection as wind and stability conditions change due to the storm (Melfi and Palm 2012). In this case, the wavelength (λ) is found by:

$$\lambda =\frac{2\pi \mid {\mathbf{V}}_{\mathbf{sh}}\mid }{N}$$ (3)

where V_sh is the boundary layer–free atmosphere wind shear, N is the Brunt-Väisälä frequency (0.008 s⁻¹ at the typical season, latitude, and altitude of the storms: Ando et al. 2012). Ruffled texture wavelengths range from 70 km on the southern margin to 24 km in the interior of this storm (typical for the interiors of storms) to 9 km on the northern margin. Thus, an increase in stability (higher N) and/or decrease in wind shear to the north could explain the change in wavelength.

On the other hand, the implied wind shear at the storm’s southern margin would be 89 m s⁻¹. Near-surface winds of 30-–40 m s⁻¹ are required to initiate saltation of sand particles on Mars, while direct lifting would require even higher winds (Read et al. 2015). A strong westerly jet is present above the boundary layer throughout most of northern fall and winter (Fig 22a-b). So it might be possible to generate high enough shear with easterly winds. However, the MCD predicts prevailing southerly or southwesterly flow near the surface (Fig 22a-b). Moreover, during the two main periods of dust storm activity, easterly winds are rare and weak while the prevailing winds are strong and broadly westerly at VL2 (Fig. 23).

Fig. 22.

MCD predictions of vertical winds and h (CBL Height) for the location and season labeled. W–E and S–N indicate zonal and meridional winds respectively and according to the usual sign convention.

Fig. 23.

Scatter plots of Viking Lander 2 observations of surface winds for the years, seasonal ranges, and the range of local times indicated. Each marker indicates an individual hourly binned wind measurement.

Gravity waves also cannot explain east–west trending ruffled texture. The MCD predicts strong westerly winds above the boundary layer during the two active periods of dust storm activity (Fig 22a-b), exactly perpendicular to what would be expected. In cases of north-south trending ruffled texture of 20–30 km wavelength (e.g., Fig. 6f), however, a role for gravity waves cannot be excluded.

Roll convection, however, can explain all of the variability in the orientation of the ruffled texture. East–west trending ruffled texture would result from strong westerlies associated with the jet mixing down to the surface. Sufficient shear in the jet will orient the rolls. The MCD suggests that background northerly winds can be stronger than westerly winds at the surface (Fig. 22b). If this were true for the extreme winds during dust storms, diagonally oriented ruffled texture (Fig. 5d) could be explained. Strong northwesterly or southwesterly surface winds would explain north-south trending ruffled texture (Fig. 6f), because these storms would experience southerly or northerly shear aloft.

Roll convection also can explain the length scales of ruffled texture. Cloud streets can be classified in terms of aspect ratio, the ratio of roll spacing to h. The aspect ratio of “classical” cloud streets is 2–4. However, cloud streets often have broader wavelengths than the classical value (Young et al. 2002; LeMone and Meitin 1984; Melfi and Palm 2012). The MCD (under CASS conditions) estimates h in the ROI to be 2.5–3.8 km when dust storms are most common (Fig. 22c). Scaling based on median dust devil heights would imply that h in the ROI varies between a few hundred meters and 3 km during northern fall and winter under relatively clear conditions (Fenton and Lorenz 2015). In addition, h might be 75% smaller under high dust opacities (Davy et al. 2009), which is consistent with the MCD under DSASS conditions (h ~ 0.75 km) (Fig 22c). Thus, the aspect ratio of ruffled texture in the storm in Fig. 10 could range from 2.6–20 (assuming the boundary layer is perturbed by the dust storm: h=3.5 km) and from 12–93 (assuming it is: h=0.75 km). The variability across the storm could be explained by the boundary layer collapsing as the storm advanced southward, which would imply the aspect ratio ranges from 12 at the northern margin (DSASS conditions) to 20 at the southern margin (CASS conditions). And this argument easily could be extended to the other storms of ruffled texture. These length scales are most consistent with “wide mixed-layer rolls”, which have aspect ratios of up to 18 (Young et al. 2002). Moreover, there is strong linear relationship between h and aspect ratio for wide mixed-layer rolls over the ocean (Young et al. 2002), but for which there is no data for h > 2.5 km.

b. Possible Triggers for Ruffled Dust Storms

Wide mixed-layer rolls on Earth are associated with cold air outbreaks over lakes or oceans (Young et al. 2002). A similar association of dust storms with post-frontal cold air advection in the ROI is plausible. The survey already excludes dust storms associated with large frontal boundaries, so all of the dust storms surveyed could be post-frontal. In addition, the location and climatology of the storms suggests some connection between the storms and northern hemisphere baroclinic wave activity. Note that storm activity is restricted to two distinct periods, one in northern fall and the other in northern winter, with a break around northern winter solstice (Fig. 3). A similar “solsticial pause” in the intensity of Mars’s northern hemisphere baroclinic wave activity and/or northern hemisphere dust storm activity has been observed by a variety of techniques (Wilson et al. 2002; Wang 2007; Lewis et al. 2016; Mulholland et al. 2016). In addition, the association of the strongest wind events at VL2 with westerly flow (Fig 23d) is potentially consistent with post-frontal cold air advection and would explain the east-west trend of much of the ruffled texture.

Moreover, cold air advection on Mars probably would generate stronger convection than another type of high wind event. The character and intensity of daytime Martian boundary layer convection depends on the near-surface superadiabatic layer (Petrosyan et al. 2011). Advecting colder air above the surface would strengthen the superadiabatic gradient and help maintain it against the boundary layer processes that counteract it: atmospheric heating by convective updrafts and absorption of infrared radiation from the surface (Petrosyan et al. 2011). Segal et al. (1997) even argued that the Martian surface and the terrestrial ocean would respond analogously to cold air outbreaks by maintaining their surface temperature against relatively large sensible heat fluxes. In the case of the ocean, this stability in surface temperature is due to its high thermal inertia. In the case of Mars, it is due to the dominance of radiative fluxes over sensible heat flux in the surface energy budget. The potentially low impact of Mars’s cold air outbreaks on surface temperature also could explain why there is no sharp boundary in surface temperature to the north of the storms, just a continuation of a similar temperature gradient to the one to the south of the storm (Fig. 13).

One theoretical explanation for the scale of wide mixed-layer rolls is latent heating (Chlond 1992; Young et al. 2002), which might be stronger in marine cold air outbreaks due to high latent heat fluxes from the ocean surface. Latent heating is minimal in the Martian atmosphere, but specific radiative heating by dust can be of comparable magnitude (Heavens et al. 2011b). Large amounts of dust mixed into the atmosphere can warm atmospheric temperatures by 10s of K (Wilson 1997; Strausberg et al. 2005; Rafkin 2009; Spiga et al. 2013; Kass et al. 2016) and high concentrations of dust at small scales might cause heating rates up to 100s of K h⁻¹ (Fuerstenau 2006; Heavens et al. 2011b). Thus, the smooth, dusty surface of the ROI supplies a potential heating source to the atmosphere analogous to the potential heating source that cold winds moving over a warm ocean supplies to the atmosphere. Yet most storms in the ROI seem to lack a warm, dust-heated core. It is possible that the dust storms are being heated by dust, but this energy either has been advected away from the storm or cannot be resolved vertically or horizontally by the observations. Interplay between dynamics and observational uncertainty is also possible. Spiga et al. (2013) simulates a storm in which there is a region of strong dust heating lying over a region of adiabatic cooling associated with convective ascent within an altitude range of ~ 5 km: a situation that could not be resolved by infrared atmospheric sounders like TES or MCS.

The ruffled storms’ restricted vertical extent and potentially cool interiors suggest that they are not thermodynamically self-sustaining. They are maintained entirely by external forcing from the hypothetical cold air advection. Once the large-scale advection weakens, the advecting airmass warms, or convection is otherwise suppressed, the storm will dissipate.

c. Other Textures

Not all storms are ruffled (Table 2). In the case of the storm of L_s = 147.53, MY 29 (Fig. 4d), its early date (Fig. 3), puffy/smoky texture, deep vertical mixing (Figs. 18 and 21a), and the unusually weak background gradient in surface temperature (Fig. 8), and deep vertical mixing suggests that its dynamics are entirely different. This storm may result from dust-heated free convection in an environment of strong surface convergence with weak wind shear and weaker stability aloft: “a rocket dust storm” in the sense of Spiga et al. (2013).

A few non-ruffled storms are too small to evaluate (Fig. 4b). Winds may have been marginal for dust lifting. Other storms have morphologies and climatology like ruffled storms but do not have ruffled texture (Fig. 4e). As these storms seem to be more common in MY 29 (when observations were at later average local time), it is possible that lifting has stopped or is much reduced from peak activity at earlier local time, resulting in breakdown of the convective rolls and merger of bands of ruffled texture. Boundary layer collapse under high dust opacity may play a role as well. The intermediate stage in this process may be shown in Figs. 6a-b and d-e.

d. Applicability to Martian Dust Storm Dynamics Generally

This study focused on the ROI mainly because several unusual-looking storms occurred there in MY 24. Indeed, the ROI itself is unusual. N. Amazonis Planitia is the part of Mars with the highest known dust devil activity as well as where the largest diameter dust devils are observed (Cantor et al. 2006; Fenton and Lorenz 2015). Dust devil activity peaks in northern summer but is minimal in northern fall and winter, a climatology roughly opposite to that of local dust storms (Cantor et al. 2006; Fenton and Lorenz 2015). Explanations for high dust devil activity here include its topographic smoothness, low altitude (high atmospheric density to lift dust), and high abundance of loose dust (Cantor et al. 2006; Fenton and Lorenz 2015).

Dust availability is probably the surface property of the ROI that most explains its unusual dust storm activity. Where dust is abundant, dust storm formation mechanisms requiring dust will operate at peak efficiency. Moreover, textural features in dust storms are tracers of the circulation (though not necessarily passive ones). If dust were not present or only mobilized in some parts of the area, the circulation might still exist but the structure of the storm would look far less organized and distinct. Smooth topography may be key as well. Topographic features will generate waves that interfere with periodic patterns generated by other processes or else generate features like lee wave clouds that may be confused with or obscure dust storm structures.

Therefore, while the ROI may be optimal for observing ruffled texture and the underlying dynamics of local dust storms generated by cold air outbreaks, both the texture and the dynamics it signifies are likely occurring elsewhere. Kulowski et al. (2016)’s multi-year survey of MOC imagery identifies plume-like texture in areas other than the ROI. If the ruffled texture was classified as plume-like in the ROI, ruffled texture may occur in the other areas as well. Moreover, the association of dust storm activity with baroclinic activity is well-known. Forcing by frontal boundaries is often considered (e.g., Cantor et al. 2001; Cantor 2007; Hinson and Wang 2010; Wang et al. 2011), but intense, post-frontal cold air advection events are likely of comparable importance.

5. Summary and Conclusions

Here was presented a study of local dust storm characteristics in a small area of Mars that is likely highly favorable for the formation of dust storms and the observation of discernible structure within them. This study is unprecedented in the number and diversity of observational datasets it applies to Martian local dust storms. The most common texture in these storms is a previously unreported ruffled texture characterized by elongated, periodic linear variations in dust opacity. These features are proposed to be structurally and dynamically analogous to the cloud streets on the Earth, particularly the wide, mixed-layer rolls that are observed over oceans and lakes during post-frontal cold air outbreaks. While the study area is optimal for the investigation of dust storm characteristics, the dynamics and even textures inferred in the area are likely not unique to it. Post-frontal cold air advection is likely an important driver of local dust storm activity wherever baroclinic activity on Mars is strong. It is also possible that ruffled texture in some storms could be explained by interactions with gravity waves.

Some storms in the area lack ruffled texture but resemble storms with ruffled texture in most other characteristics. They therefore may be due to weaker outbreak events and/or represent the decay phase of ruffled storms. One storm in the area occurs unusually early in the year, mixes dust unusually deeply, and has a puffy/smoky texture. This storm likely developed in an environment of strong surface convergence with minimal vertical shear as opposed to the strong vertical shear environment implied by the ruffled storms. All of these storms cool the surface and most appear to have a negative or neutral impact on atmospheric temperature, implicating them as cold core weather systems that must be sustained by external forcing.

This study therefore has three implications for atmospheric studies on Mars and one for atmospheric science in general. First, it identifies a class of dust storms with interesting textures that form in an area with relatively simple and idealizable surface topography and thermophysical properties. These storms are therefore ideal targets for mesoscale simulations and idealized models of dust storm structure. These simulations also could confirm or invalidate the analogies between terrestrial and martian atmospheric dynamics upon which this analysis has often relied. Second, this study proposes a tentative large-scale dynamical mechanism for these storms, which can be tested against re-analysis datasets for Mars. Third, it demonstrates that a variety of interesting aspects of dust storm dynamics are confined to the first atmospheric scale height, necessitating the development of improved observational techniques to better understand them. Finally, this work connects a type of dust storm on Mars with cloud streets on Earth, raising further questions about why the aspect ratios of some rolls greatly exceed the values predicted by classical theory but providing a set of new test cases for testing hypotheses old and new.