Modeling the Martian dust cycle and surface dust reservoirs with the NASA Ames general circulation model

Authors


Abstract

[1] We employ the NASA Ames Mars general circulation model (GCM) to investigate the dust lifting mechanisms responsible for the observed Martian dust cycle and the net surface response to the combined influence of dust lifting and deposition. This GCM includes lifting, transport, and deposition of radiatively active dust. Two dust lifting mechanisms are accounted for: wind stress lifting and dust devil lifting. A “baseline” simulation is presented and shown to compare well to available spatial and temporal observations of atmospheric opacity, wind stress dust lifting events, and atmospheric temperatures recorded during a nonglobal dust storm year. Multiple simulations were conducted to explore the model's sensitivity to a wide range of dust lifting parameters (the functional dependence of surface dust flux on wind stress, the wind stress threshold for lifting, etc.) Model results robustly suggest that wind stress lifting produces the peak in atmospheric dust load during southern spring and summer and that dust devils maintain the background haze of atmospheric dust during northern spring and summer. These results are consistent with previously published conclusions. Dust devil and wind stress lifting contribute equally to the simulated total amount of dust lifted annually during nonglobal dust storm years. The simulated spatial pattern of annual net deflation/deposition suggests that the low thermal inertia regions (Tharsis, Arabia, and Elysium) are not currently net dust accumulation regions. This net deflation is the result of dust devil dust lifting, suggesting that dust devils could play an important role in the present-day pattern of surface dust reservoirs.

1. Introduction

[2] Aeolian processes dominate the current surface environment on Mars [Leovy, 2002; Armstrong and Leovy, 2005]. Evidence of recent dust lifting, transport and deposition is abundant through observations of atmospheric dust opacity [Smith, 2004], dust devil tracks [Fisher et al., 2005] and albedo [Christensen, 1988; Smith, 2004; Kahre et al., 2005a].

[3] Airborne dust plays a critical role in the thermal and dynamical state of the Martian atmosphere. Suspended dust particles absorb solar radiation and absorb and emit infrared radiation. Atmospheric heating rates are strongly affected by atmospheric dust [Gierasch and Goody, 1968]. Atmospheric heating drives dynamical processes; thus the atmospheric dust load strongly influences the circulation [Haberle et al., 1982].

[4] The Martian atmosphere contains dust throughout the year, but the atmospheric dust load varies with season. The peak of the annual atmospheric dust optical depth occurs during southern spring and summer when Mars is near perihelion [Smith, 2004]. During these seasons, regional storms persist throughout the southern hemisphere and northern baroclinic zone. While the dust activity is heightened during southern spring and summer, a background haze of atmospheric dust is maintained throughout northern spring and summer. Mars Global Surveyor (MGS) Mars Orbiter Camera (MOC) observations indicate that small regional dust events persist throughout northern spring and summer, but they are less efficient at increasing the overall atmospheric dust load during northern spring and summer [Cantor et al., 2001]. On the basis of observations [Murphy and Nelli, 2002; Fisher et al., 2005] and general circulation modeling results [Basu et al., 2004], it is likely that dust devils are responsible for the presence of atmospheric dust during these seasons.

[5] Dust devils were first identified in Viking lander meteorology data [Ryan and Lucich, 1983] and Viking Orbiter images [Thomas and Gierasch, 1985]. More recently, dust devils and dust devil tracks have been found to be widespread in Mars Orbiter Camera images [Edgett and Malin, 2000; Balme et al., 2003; Fisher et al., 2005]. Balme et al. [2003] counted dust devil tracks and determined that the quantity of dust lifted by dust devils is insufficient to maintain the background atmospheric dust haze. Fisher et al. [2005] counted both dust devils and dust devil tracks in their survey and reached the conclusion that the quantity of dust lifted by dust devils could maintain the background atmospheric dust haze. Dust devil activity has been found to peak during local summer [Fisher et al., 2005]. Of the nine regions surveyed by Fisher et al. [2005], Amazonis contained the highest frequency of dust devil occurrence. We use these dust devil observations to constrain the dust devil lifting parameterizations that are implemented in the numerical model simulations described in this work.

[6] The observed Martian dust cycle depends upon the availability of mobile dust particles on the surface. Surface dust deposits are known to exist on Mars in at least the three low thermal inertia regions of Arabia, Tharsis, and Elysium [Christensen, 1986]. Based on observations of thermal inertia and rock abundance, these deposits are thought to be between 0.1 and 2 m deep [Christensen, 1986]. Observational estimates of deposition rates in these regions range from a few to almost 50 μm yr−1 and tend to be based on two assumptions: (1) after a global dust storm, dust that is deposited in the northern hemisphere is done so uniformly [Christensen, 1988]; and (2) once dust is deposited in the low thermal inertia regions, it is not subsequently removed [Christensen, 1988]. Christensen [1986] estimated an age of between 105 and 106 years for these regions based on the estimated thickness of the low thermal inertia deposits and an estimated rate of deposition. Since this timescale is similar to the period of orbital oscillations, it was postulated that these deposits are young and exhibit a cyclic behavior of deposition and then deflation [Christensen, 1986].

[7] Investigations of the Martian dust cycle with numerical general circulation models have to date varied greatly in project focus. Newman et al. [2002] focused on present-day Mars. They studied the effects of implementing fully interactive dust (i.e., dust lifting and radiatively active dust transport) into the Oxford/LMD GCM. Two dust lifting mechanisms were employed: dust devil and wind stress lifting. Simulated wind stress lifting peaked during southern summer. Simulated dust devil lifting peaked during southern summer and had a secondary peak during northern summer.

[8] Haberle et al. [2003] addressed the history of the three low thermal inertia regions by analyzing the pattern of model predicted wind stress lifting over a wide range of obliquity. They found that these regions do not experience large amounts of wind stress deflation at any obliquity. However, their model did not include dust devil lifting or dust transport and deposition. As we will show, dust devil lifting plays a dominant role in determining the net dust gain or loss in these low thermal inertia regions.

[9] Basu et al. [2004] used a fully interactive version of the Geophysical Fluid Dynamics Laboratory (GFDL) GCM to study current-day Martian climate. Dust devil lifting was shown to be the most likely mechanism for maintaining the background haze of atmospheric dust during northern spring and summer when wind stress lifting is minimal. The relative roles of dust devil and wind stress lifting on the spatial pattern of net dust deflation and deposition was not discussed. Such a discussion is included in this work.

[10] Newman et al. [2005] simulated the dust cycle under a range of orbital configurations with both radiatively active and radiatively inert (i.e., passive) dust transport. While dust devil lifting was included in their passive simulations, it was not included in their radiatively active simulations. On the basis of a passive simulation, they found the current dust deposition rate in one of the low thermal inertia continents, Arabia, to be approximately 1.5 μm yr−1. In this work, simulations that include radiatively active dust transport are used to study the net dust deposition/deflation rates in these low thermal inertia regions.

[11] The goal of this work is to increase our understanding of the physical processes that control the current Martian dust cycle and the depth evolution of the dust deposits in the low thermal inertia continents. Wind stress and dust devil lifting parameterizations are included in the NASA Ames Mars general circulation model to explore the respective roles of these dust lifting mechanisms on the dust cycle and on the spatial pattern of net surface dust deflation and deposition. We present a “best fit” baseline simulation that most closely reproduces available observations of atmospheric opacity, wind stress dust lifting events, and atmospheric temperatures during a nonglobal dust storm year. However, since many uncertainties exist in our understanding of the physical processes that drive the Martian dust cycle, we recognize that our baseline simulation does not capture all aspects of the “real” Mars. Therefore a large range of dust lifting parameters have been implemented in order to understand the model sensitivities upon the pattern of net surface dust deflation and deposition.

[12] The simulations presented here do not reproduce the interannual variability of global dust storms. We do not believe that the physical processes that drive interannual variability are well understood. Basu et al. [2004] simulate interannual variability with high wind stress lifting thresholds. As we show in section 4.2, when comparably high wind stress thresholds are implemented into the NASA Ames GCM, the spatial extent of observed dust lifting events [Cantor et al., 2001] is not well reproduced. Additionally, little discussion of the nature of the simulated interannual variability is presented by Basu et al. [2004], making direct comparisons between their results and ours difficult. Therefore we have chosen to not focus on interannual variability in this current work.

[13] Each simulation presented will be given an identifier; details concerning all simulations are listed in Table 1. Section 2 describes the model and the treatment of dust in the model. Section 3 presents the results from the baseline simulation which will be shown to be the best fit to the observed dust cycle. Multiple dust lifting schemes and their associated parameters will be explored in section 4. Section 5 describes the results of sensitivity studies performed to determine the robustness of the spatial pattern of net surface dust deposition and deflation.

Table 1. All Simulations
IdentifierWind Stress Schemeτ*, mN m−2αWDust Devil SchemeαD, kg J−1
KMH0225INDEPKMH22.50.1independent10−10
KMH01INDEPKMH100.02independent10−10
KMH035INDEPKMH350.45independent10−10
KMH05INDEPKMH503.0independent10−10
NEW01INDEPNewman102.25 × 10−6 m−1independent10−10
NEW0225INDEPNewman22.59.0 × 10−6 m−1independent10−10
NEW035INDEPNewman354.0 × 10−5 m−1independent10−10
NEW050INDEPNewman503.5 × 10−4 m−1independent10−10
DDEVONLYINDEPindependent10−10
DDEVONLYDEPdependent4.0 × 10−10
KMH0225DUSTYKMH22.50.15independent10−10
KMH0225NODEVKMH22.50.1

2. NASA Ames Mars GCM

2.1. General Model Description

[14] The NASA Ames Mars GCM (Version 1.7.3) is a three-dimensional (3-D) grid point model of the Martian atmosphere. The primary difference between this version and the version used for recently published NASA Ames GCM studies [Haberle et al., 1999, 2003] is the numerical grid. Additionally, a tracer transport scheme is incorporated in GCM 1.7.3. A manuscript describing GCM 1.7.3 in detail is in preparation. Therefore only a summary of the model is given here, with emphasis on the processes important for a fully interactive dust cycle.

[15] GCM 1.7.3 runs on an Arakawa C-Grid; the resolution used here is 5° in latitude by 6° in longitude. The normalized vertical pressure sigma coordinate system increases in resolution progressively as altitude decreases. Near the surface the vertical resolution is on the order of tens of meters, while above 10 km the vertical resolution is approximately one half of one scale height. Surface properties include Mars Observer Laser Altimeter (MOLA) topography smoothed to a 5° by 6° resolution, and the albedo and thermal inertia maps that are currently being used in the Laboratoire de Meteorologie Dynamique (LMD) model (Forget, personal communication). The thermal inertia maps consist of both Viking and Thermal Emission Spectrometer (TES) data. The TES data are used for tropical and midlatitudes [Jakosky et al., 2000], and Viking data are used for the polar regions [Paige et al., 1994; Paige and Keegan, 1994]. Both the TES and Viking data sets have been corrected for the effects of atmospheric dust [Haberle and Jakosky, 1991].

2.2. Planetary Boundary Layer Scheme

[16] Dust is lifted through interactions between the surface and the atmosphere that occur at the bottom of the planetary boundary layer (PBL). The boundary layer scheme employed in this version of the model was adapted from Haberle et al. [1999]. Simplifications were made to the Haberle et al. [1993] model to increase its numerical stability; these modifications had little effect on the simulated fields [Haberle et al., 1999]. The equations that describe the heat (H) and momentum (τ) fluxes are

equation image
equation image

where u* and θ* are related to the mean wind u and the air-to-ground temperature gradient ΔΘ, respectively, in the following way:

equation image
equation image

Cdm and Cdh are the drag coefficients of momentum and heat, respectively.

[17] The friction drag coefficients depend upon the surface roughness (z0), the near surface vertical wind speed gradient, and the near surface vertical thermal gradient. The spatial variation of the surface roughness is not well known on Mars, so previous GCM studies have assumed a spatially uniform value of 1 cm. This spatially uniform value for z0 is employed in all of the simulations presented here.

[18] Further details on the planetary boundary layer formulation used in this version of the GCM are given by Haberle et al. [1999].

2.3. Dust in the Model

2.3.1. Particle Size Distribution

[19] The model contains interactive dust (i.e., surface lifting, atmospheric transport and gravitational sedimentation). This includes the spatially and temporally varying suspended dust opacity that interacts with both visible and infrared radiant energy.

[20] The physical properties of airborne dust have been constrained using remotely sensed visible and IR spectral measurements from Mariner 9 [Pang et al., 1976; Chylek and Grams, 1978], Viking lander [Pollack et al., 1979, 1995], Viking Orbiter Infrared Thermal Mapper (IRTM) emission phase functions (EPF) [Toon et al., 1977; Clancy and Lee, 1991], Pathfinder [Tomasko et al., 1999] and MGS TES [Clancy et al., 2003; Wolff and Clancy, 2003] data. Key dust parameters include the particle size distribution, particle shape, and the size-dependent extinction coefficient. Particle size distributions have been described using gamma distributions [Tomasko et al., 1999], modified gamma distributions [Toon et al., 1977], and lognormal distributions [Pollack et al., 1995]. Recent analysis indicates that the properties of airborne dust particles vary in space and time and have an effective radius ranging from 1.0 μm to 2.5 μm [Clancy et al., 2003; Wolff and Clancy, 2003].

[21] The dust mass lifted from the surface in the model by the parameterizations described below is distributed in a lognormal particle size distribution that produces an average particle size of radius 1.5 μm. The distribution is represented by three particle sizes of radii 0.1 μm, 1.5 μm and 10 μm. For a given lifted dust mass, 3.0 × 10−5% of this mass (1% of the opacity) is contained in the 0.1 μm particle size bin, 19.8% (77% of the opacity) in the 1.5 μm particle size bin, and 80.2% (22% of the opacity) in the 10 μm particle size bin. While the lifting parameterizations lift dust mass, the quantity that is important for the radiative transfer calculations is the optical depth. This distribution yields 6.55 × 10−4 g cm−2 of dust per unit optical depth. This relationship between mass and optical depth changes as the suspended particle size distribution evolves. The mass per unit optical depth would be decreased by a factor of approximately 3.5 if all of the 10 μm particles fall out of the atmosphere, leaving the two smaller particle sizes accounting for all of the optical depth.

[22] Using this same lognormal distribution, various simulations were designed to test the sensitivity of the number of particles used to represent the distribution on the outcome on the simulated dust cycle. Using 10 particle sizes, the distribution gives 2.9 × 10−4 grams per optical depth in the visible, which is more than a factor of 2 less than the 3 particle mass per optical depth value. Because of this difference in opacity lifted per unit dust mass lifted, the atmospheric dust load increased with 10 particle sizes. In spite of this difference, the simulated dust cycle in the two simulations was quite similar.

2.3.2. Radiative Transfer

[23] The model radiative transfer scheme accounts for the radiative effects of gaseous CO2 and suspended dust [Pollack et al., 1990; Haberle et al., 1999]. At visible wavelengths, the radiative effects of CO2 and suspended dust are used to calculate the net solar flux at each vertical layer. At infrared wavelengths, CO2 and suspended dust are accounted for in two spectral intervals: one covering the 15-μm vibrational CO2 band and one covering the spectral region outside the 15-μm band. The single scattering albedo employed in the model is 0.86 [Pollack et al., 1979, 1990], which is low based on more recently derived values from TES [Wolff and Clancy, 2003; Clancy et al., 2003]. The asymmetry parameter is 0.79 [Pollack et al., 1979, 1990], and the ratio of the visible to infrared opacity (absorption only) is assumed to be 2.

2.3.3. Horizontal and Vertical Transport

[24] Dust in the atmosphere is transported horizontally by model resolved winds. GCM 1.7.3 has a second-order tracer transport scheme incorporated into the dynamical core [Suarez and Takacs, 1995]. In addition to horizontal transport, vertical transport is included in the form of turbulent mechanical mixing (eddy mixing), diagnosed vertical winds, and gravitational settling. The dust particle fall velocity is calculated via the Stokes-Cunningham relationship for each individual size particle bin; it is then converted into a sedimentation flux through each model layer. The 10 μm particle falls approximately 100 faster than the 0.1 μm particle and 10 times faster than the 1.4 μm under typical near-surface Martian conditions.

2.3.4. Surface Dust Lifting Parameterizations

[25] In order for dust to enter into the atmosphere and affect the thermodynamic consequences described above, the dust must first be “injected” from the surface into the atmosphere. The model includes parameterizations for wind stress dust lifting and dust devil lifting, which depend on the momentum and heat exchanged between the atmosphere and the surface.

2.3.4.1. Wind Stress Lifting

[26] Observations of local and regional dust lifting events indicate that the momentum imparted to the surface by winds is one mechanism responsible for lifting dust off the Martian surface [Cantor et al., 2001]. While wind speeds on Mars are not predicted to be strong enough to lift dust sized particles directly [Greeley and Iversen, 1985], simulated near-surface wind speeds do exceed the experimentally derived threshold for sand-sized particles to enter into saltation [Greeley and Iversen, 1985]. Previous GCM studies have implemented parameterizations for wind stress lifting [Newman et al., 2002, 2005; Basu et al., 2004].

[27] The wind stress lifting scheme employed here (hereafter referred to as the “KMH” scheme) was originally formulated by Westphal et al. [1987] for Earth-based dust lifting. Westphal et al. used two- and three-dimensional numerical models to simulate Saharan dust storms by implementing a surface dust lifting rate prescription that was based on terrestrial observations. This prescription for vertical dust mass flux (Fd) was formulated in terms of the model-predicted threshold friction velocity (u*), below which the lifting rate was zero. The Westphal et al. formulation was modified to be prescribed in terms of the surface stress due to the significant range of Martian topographic elevation and thus surface gas density. This modification allows for a spatially uniform stress threshold. Finally, appropriate alterations were made for Martian gravity. The magnitude of the vertical surface dust flux (FW) is prescribed as

equation image

where τ* is the specified threshold stress required for lifting.

[28] While this lifting scheme injects dust mass into the atmosphere, the important quantity for the radiative effects of suspended dust is opacity. Therefore this lifting scheme is sensitive to the particle size distribution employed. Haberle et al. [2003] used this lifting scheme to calculate the surface dust deflation potential with a version of the NASA Ames Mars GCM that was driven by a spatially and temporally fixed atmospheric dust load. A stress threshold of τ* = 22.5 mN m−2 was employed. This value was based upon an experimentally derived stress threshold (35 mN m−2 [Greeley and Iversen, 1985]) multiplied by 0.8 to modify it for a dynamic state. An additional factor of 0.8 was added to account for the coarse model resolution which results in reduced average wind velocities. This stress threshold was verified in fully interactive dust cycle simulations [Murphy, 1999]. Because the atmospheric dust load was fixed, no assumption needed to be made about the lifted particle size distribution [Haberle et al., 2003]. With the particle size distribution described above and a higher-resolution version of the model, using this prescription as written in our current fully interactive version of the GCM yields unrealistically large atmospheric dust loads. Therefore a dimensionless efficiency parameter (αW) is added as a leading constant, which allows the wind stress lifting prescription to be “tuned” to match observations. As an example of the sensitivity of the dust lifting rate to the wind stress, when a wind stress threshold of 22.5 mN m−2 is employed, the lifting rate for τ = 26.9 mN m−2 is twice the lifting rate for τ = 25 mN m−2.

[29] Within this work, this KMH wind stress lifting scheme is compared in detail to the scheme developed by Newman et al. [2002, 2005]. The Newman scheme is based on a formulation of the vertically integrated horizontal sand particle flux that was developed by White [1979]. Assuming that the vertical dust flux lifted by near-surface wind stress was proportional to this horizontal dust flux, Newman et al. [2002] developed a formulation with a constant of proportionality (called the “efficiency factor”) to relate lifted dust flux to friction velocity [Newman et al., 2002]. In its most recent form, the Newman et al. [2005] parameterization is functionally dependent upon the surface stress:

equation image

where g is the gravitational acceleration at the surface, ρ is the density of the atmosphere near the surface, and αN is an efficiency factor with units of m−1. When this scheme is used with a threshold of 22.5 mN m−2, the lifting rate for τ = 27.4 mN m−2 is twice the lifting rate for τ = 25 mN m−2.

[30] While both the KMH and Newman wind stress lifting schemes depend on surface stress, the functional dependence of the dust lifting rate on this quantity differs. The lifting rate as a function of surface stress for the Newman scheme (assuming ρ = 0.01 kg m−3) and the KMH scheme for several different pairs of threshold stress and efficiency factors are compared (Figure 1). All four of the tuned (i.e., the efficiency factor was changed to produced a reasonable dust cycle) threshold stress and efficiency factor pairs employed in our wind lifting scheme that will be discussed in sections 3 and 4 are plotted. Additionally, the threshold stress and efficiency factor pair (τ* = 10 mN m−2, αW = 2.25 × 10−6 m−1) that were used for radiatively active simulations by Newman et al. [2005] are plotted. Finally, the Newman scheme was incorporated into and employed in the Ames GCM with threshold stresses of 22.5, 35, and 50 mN m−2 and tuned in each case to produce a reasonable dust cycle; these pairs of parameters are also included. The Newman scheme lifts more dust at low surface stress values but less at high surface stress values than the KMH wind stress lifting scheme. Comparisons of the dust cycle results from both the KMH scheme and the Newman scheme are presented in section 4.2.2.

Figure 1.

Comparison of the lifting rate (kg m−2 s−1; logarithmic scaling below the dotted line and linear scaling above the dotted line) versus surface stress (mN m−2) for tuned (using the NASA Ames Mars GCM) simulations with the KMH and Newman wind stress lifting schemes for τ* = 10, 22.5, 35, and 50 mN m−2. A near-surface air density of 0.01 kg m−3 was assumed for the Newman scheme.

2.3.4.2. Dust Devil Lifting

[31] Previous GCM simulations incorporating only resolved wind stress lifting that were tuned to match the observed opacity peak during southern spring and summer do not produce the observed background dust haze during northern spring and summer [Newman et al., 2002; Basu et al., 2004]. Therefore another dust lifting mechanism is needed. The contribution of dust devil lifting to the annual dust cycle has been shown to be significant through both observations [Fisher et al., 2005; Balme et al., 2003] and numerical modeling [Newman et al., 2002; Basu et al., 2004]. Two dust devil lifting schemes were implemented by Newman et al. [2002] into the Oxford/LMD GCM; those two schemes are compared in this current work by incorporating them into the Ames GCM. Both schemes depend on the heat exchange between the surface and the atmosphere. The fundamental difference between these schemes is that one scheme is independent of a predicted threshold tangential velocity and one scheme depends on a threshold tangential velocity for lifting.

[32] The threshold-independent scheme is based on the thermodynamics of dust devils studied by Rennó et al. [1998] and was first used in a GCM by Newman et al. [2002]. On the basis of the magnitude of the vertical surface-to-atmosphere heat flux and the depth of the planetary boundary layer (both quantities are prognostic products of the PBL scheme), the lifting prescription is defined as

equation image

Fs is the vertical sensible heat flux (W m−2), αD is a tunable efficiency parameter with units of kg J−1, and b is defined as

equation image

where ps is the surface pressure, pcon is the pressure at the top of the PBL, and χ is the specific gas constant divided by the specific heat capacity at constant pressure. This parameterization allows dust to be lifted whenever the vertical heat flux is positive; therefore dust devil lifting occurs at most locations throughout daylight hours except where surface CO2 ice is locally present.

[33] A threshold-dependent dust devil lifting scheme was also used by Newman et al. [2002]. This parameterization is based on a semiempirical formula for the threshold vortex tangential wind speed required for lifting dust found in laboratory dust devil experiments [Greeley and Iversen, 1985]. The threshold tangential wind speed depends on the surface air density, the dust particle density, and the dust particle diameter. The tangential wind speed depends on the pressure drop from the perimeter to the center of the vortex, which is related to the height of the planetary boundary layer [Rennó et al., 1998]. The dust mass flux lifted by dust devils depends on the tangential wind speed (vt) [Newman et al., 2002]:

equation image

where αD is a tunable efficiency factor with units of kg J−1, ρ is the near-surface air density, and g is Martian gravity.

[34] The total dust lifted by either dust devil lifting schemes in one simulated nonglobal dust storm year is comparable in magnitude to the total dust lifted by either wind stress lifting scheme. This indicates that although the simulated dust devil lifting rates are small compared to the simulated wind stress lifting rates, the spatially and temporally integrated effect of dust devils is important to the overall dust cycle.

3. Baseline Simulation Validation and Results

[35] The goal of this work is to increase our understanding of the physical processes that control the Martian dust cycle. To do this, a set of dust lifting parameters was chosen to produce the best fit to observations of atmospheric dust opacity and dust lifting events during a nonglobal dust storm year. This simulation is referred to as the “baseline” simulation; one year of this simulation is described in detail in this section. The baseline simulation was also integrated for multiple Martian years. Results of that simulation were presented by Kahre et al. [2005b] and will be discussed briefly in section 3.1. Comparisons of this baseline simulation with other simulations will be presented in section 4 in order to explore the physical processes and model sensitivities to the lifting parameters chosen for the baseline simulation. The baseline simulation includes the KMH wind stress lifting prescription with a stress lifting stress threshold of 22.5 mN m−2 and an efficiency factor (αW) of 0.1. The threshold-independent dust devil lifting prescription [Newman et al., 2002] was included with an efficiency factor (αD) of 1 × 10−10 m−1.

[36] The model was initially run from isothermal static conditions for one Martian year with the baseline lifting parameters in order to bring the CO2 cycle into equilibrium. All simulations presented were initiated approximately 80 sols (Martian days) before the end of this spin-up year, which was enough time for each simulation to reequilibrate before the new year started. It was assumed that there was a planet-wide infinite surface dust reservoir. Dust could potentially be lifted as long as CO2 ice was not present on the surface.

3.1. Atmospheric Dust Loading

[37] The pattern of zonally averaged 9 μm opacity (scaled to the 6.1 mbar pressure surface) generated in this baseline simulation agrees reasonably well with TES observations during Mars years 24 and 26 (Figure 2; “Mars year” numbering was defined by Clancy et al. [2000]; Mars year 1 began on 11 April 1955 at Ls = 0; Mars year 25 began on 1 June 2000). During northern spring and summer, 9 μm opacities are on the order of 0.1 and are generally spatially and temporally constant. As will be shown below, this background dust haze is due to dust devil lifting. As southern spring is approached, regional storms (wind stress dust lifting events) increase the local opacity in regions in the southern midlatitudes and near the north pole. The peak zonally averaged opacities occur near Ls = 285. Two spatial maxima are apparent at this season in the zonally averaged opacity, one in the south and one (that is smaller in magnitude) in the north. The dust lifting sources for these maxima will be described in section 3.3.6. Major features that are in agreement between the simulated zonally averaged opacity pattern and the TES-derived opacity pattern are the spatially extensive low dust regime that exists during northern spring and summer and the increase in suspended dust abundance during southern spring and summer. The largest difference between the model results and observations is the season of the peak opacity. TES observations of Mars year 24 show the peak in the zonally averaged opacity to be tens of degrees of Ls earlier than the modeled peak opacity. TES-derived zonally averaged opacity of Mars year 26 show the peak in the zonally averaged opacity to be tens of degrees of Ls later than the modeled peak opacity. As will be discussed in sections 4.2.1 and 4.2.2, the peak opacity occurs at approximately the same Ls when any combination of tuning parameters is used when the magnitude of the opacity reproduces the observed dust cycle during a nonglobal dust storm year. When the baseline simulation is integrated for multiple years, some year-to-year variation does occur in the magnitude of the peak globally averaged opacity and the season of Ls [Kahre et al., 2005b]. However, this multiyear baseline simulation does not produce spontaneous global dust storms [Kahre et al., 2005b].

Figure 2.

(top) Zonally averaged 9 μm opacity for the baseline simulation. (middle and bottom) TES-derived zonally averaged 9 μm opacity for Mars years 24 and 26, respectively. Opacities have been scaled to the 6.1 mbar pressure level.

[38] As expected on the basis of the pattern of the zonally averaged opacities, the globally averaged opacity is also in agreement with the TES-derived globally averaged opacity for Mars years 24 and 26 (Figure 3). The minimum of the simulated and observed globally averaged opacity occurs during northern spring and summer, and the peak simulated and observed globally averaged opacity occurs during southern spring and summer. Once again, the Ls of the simulated peak globally opacity occurs in between the Ls values of the TES-derived peak opacity for Mars years 24 and 26.

Figure 3.

Globally averaged 9 μm opacity from the baseline simulation. TES-derived globally averaged 9 μm opacity from Mars years 24 and 26. Opacities have been scaled to the 6.1 mbar pressure level. Simulated globally averaged lifting rates are due to dust devils and wind stress dust lifting.

3.2. Atmospheric Thermodynamics

[39] The magnitude and vertical structure of the model-predicted temperatures compare well to TES-derived temperatures during the solstices and equinoxes (Figure 4) [Banfield et al., 2004]. At both equinoxes, the zonally averaged temperature field is equatorially symmetric, with peak temperatures near the surface at the equator. The temperature decreases as the pole is approached; minimum temperatures occur at altitude over the poles. During these equinox seasons, two reasonably symmetric Hadley cells dominate the circulation. Westerly winds exist in the midlatitudes of both hemispheres, and easterly winds blow in the tropics. At the solstices, the maximum temperatures occur at high latitudes in the summer hemisphere and the minimum temperatures occur over the winter pole. The circulation consists of a single cross-equatorial Hadley cell with the rising branch in the summer hemisphere near 30° in latitude and the descending branch in the winter hemisphere near 30° in latitude. During these seasons, easterly winds exist throughout the summer hemisphere while westerly winds exist throughout the winter hemisphere. Dust reaches its maximum mixing ratio and is mixed to the highest altitudes during southern summer, which corresponds to the season during which the Hadley cell is the most intense.

Figure 4.

(left) Zonally averaged mass stream function and vertical dust mixing ratio for Ls = 0, 90, 180, and 270 (top to bottom). (right) Zonally averaged temperature and zonal winds for Ls = 0, 90, 180, and 270 (top to bottom).

[40] Dust lifting due to winds is caused by the momentum exchange between the atmosphere and the surface. The parameter used to describe this process is the surface stress (τ), which depends upon the near surface wind speed and its vertical gradient, the near surface vertical temperature gradient, and the near surface air density. The temporally varying spatial pattern of surface stress is therefore directly connected to the locations of simulated wind stress lifting (Figure 5). Regions of increased surface stress during all seasons include Hellas and Tharsis, but the surface stresses in these regions are the highest during southern summer. The regions of maximum surface winds correspond to regions of maximum surface stress and maximum wind stress lifting rates. This suggests that, at surface stresses that exceed the threshold stress, wind speed dominates over the thermal gradient.

Figure 5.

(left) Ten-sol average map of surface winds for Ls = 0, 90, 180, and 270 (top to bottom). (middle) Ten-sol average map of surface stress for Ls = 0, 90, 180, and 270 (top to bottom). (right) Ten-sol average map of wind stress lifting rate for Ls = 0, 90, 180, and 270 (top to bottom). Many of the regions that experience increased surface winds, surface stresses, and wind stress lifting rates are colocated with topographical features. The reader is referred to Figure 7 (bottom) for the model's topography.

3.3. Surface Dust Reservoir

[41] The evolution of surface dust reservoirs depends on the spatial and temporal patterns of both dust lifting and deposition. Including dust transport in the GCM allows dust lifting and deposition patterns to be analyzed.

[42] During southern spring and summer, regions of increased wind stress lifting occur along the periphery of the receding south polar cap, in localized regions throughout the southern midlatitudes (specifically Hellas and Argyre), along the periphery of the growing north polar cap, and throughout the baroclinic zone in the northern midlatitudes (Figure 6, top). During southern autumn and winter, wind stress lifting occurs along the periphery of the growing south polar cap, along the edge of the receding north polar cap, and in localized regions in the northern midlatitudes (slopes of the Tharsis volcanoes) (Figure 6, top). Compared to the regionally confined wind stress lifting, dust devils provide a spatially ubiquitous dust source. Dust devil lifting peaks in the north during northern summer and in the south during southern summer (Figure 6, middle).

Figure 6.

Zonally averaged dust lifting rate (kg m−2 sol−1) due to (top) wind stress and (middle) dust devil lifting and (bottom) dust deposition rate (kg m−2 sol−1) from baseline simulation. Red contours highlight regions of maximum dust lifting/deposition rates. Heavy black lines outline the north and south polar cap edges.

[43] The dust mass deposition pattern is similar to the lifting pattern (Figure 6, bottom), which indicates that the majority of the lifted dust mass falls back to the ground before it is transported away. This is to be expected since 80% of the lifted dust mass goes into the 10 μm particles that do not travel far before falling back to the surface. The dust deposition pattern of the smaller particles [not shown] does not reflect the lifted pattern, suggesting that they are transported away by winds.

[44] The pattern of net surface dust deflation (shaded solid contours) and deposition (white dotted contours) over one Martian year is shown in Figure 7. Regions of net deposition include the center of Hellas, Noachis Terra, Sinai and Solis Planum, the north and south polar regions, Acidalia, Chryse, and regions in between the Tharsis volcanoes. Regions of substantial net deflation include the northeast and southwest interior slopes of Hellas and the slopes of the Tharsis volcanoes.

Figure 7.

(top) Map of net annual dust deflation (shaded regions) and deposition (nonshaded regions) for the baseline simulation. The heavy black boxes outline the three low thermal inertia regions: Tharsis (left), Arabia (middle), and Elysium (right). (bottom) MOLA topography smoothed to the model's resolution. Contours indicate height above (solid) or below (dotted) the MOLA geoid in meters. The contour interval is 1000 m.

[45] Net dust deposition occurs in both the north and south polar regions throughout the year (Figure 8). The dust deposition rate in the south polar region remains approximately constant in time at an annually integrated rate of 2.1 μm yr−1 (5.25 × 10−3 kg m−2 yr−1). The dust deposition rate in the north polar region is seasonally dependent, with a very small deposition rate (i.e., approximately 0.9 μm yr−1) during northern spring and summer and an increased rate (i.e., approximately 5.9 μm yr−1) during southern spring and summer. Averaged over the year, the deposition rate in the northern polar region is approximately 3.1 μm yr−1 (7.75 × 10−3 kg m−2 yr−1). At the season of maximum cap mass in the north and the south, the seasonal caps contain 620 kg m−2 and 1070 kg m−2 of CO2, respectively. On the basis of this simulation, the north and south seasonal caps contain approximately one part per million of dust.

Figure 8.

Net surface dust depth deflated (negative) or deposited (positive) as a function of Ls in the north and south regions for the baseline simulation.

[46] All three of the low thermal inertia continents (i.e., Arabia, Tharsis and Elysium; see boxes in Figure 7) contain large regions of localized simulated annual net surface dust deflation. When averaged over the total area of each low thermal inertia continent, the net deflation/deposition rates can yield spatially averaged information about these regions. While Elysium experiences net spatially averaged deposition, Arabia and Tharsis experience net spatially averaged deflation. However, the net spatially averaged dust deflation rates in these locations are quite small; only very few locations (i.e., regions on the slopes of the Tharsis volcanoes) experience a net deflation rate in excess of 1 μm yr−1 (Table 1). The spatially averaged time histories of net surface dust deposition/deflation in Arabia, Tharsis, and Elysium (Figure 9) indicate that these regions experience deflation during northern spring and summer and deposition during northern autumn and winter.

Figure 9.

Net surface dust depth deflated (negative) or deposited (positive) as a function of Ls in the three low thermal inertia regions for the baseline simulation.

4. Sensitivity to Dust Lifting Parameters

[47] Since both the dust devil and wind stress dust lifting mechanisms play significant roles in the baseline simulated dust cycle and net deposition and deflation results, it is worth exploring these lifting processes more thoroughly. The two dust lifting mechanisms, wind stress and dust devils, were tested individually as well as in combination with a large range of parameters. When only wind stress lifting was implemented (with the baseline wind stress threshold and efficiency factor), the observed peak annual opacity was reproduced, but the atmospheric dust load dropped to nearly zero during northern summer (Figure 10), which was inconsistent with observations. When only the dust devil lifting scheme is employed, the observed background opacity (globally and zonally averaged) throughout northern summer was well reproduced, but the southern autumn and summer opacities were much too low (Figure 10). These results are consistent with previous model results [Basu et al., 2004; Newman et al., 2002]. Further sensitivity studies of the dust devil and wind stress lifting schemes were performed in order to justify the parameters and/or lifting schemes used for the baseline simulation.

Figure 10.

Globally averaged 9 μ opacity for a simulation with only wind stress lifting (solid curve) and a simulation with only dust devil lifting (dotted curve).

4.1. Dust Devil Parameterizations

[48] Two dust devil lifting schemes are compared, the threshold independent (used in the baseline simulation discussed above) and threshold-dependent lifting schemes described above and used by Newman et al. [2002] (Table 2). Simulations were run without wind stress lifting. In both the threshold-independent and threshold dependent simulations, dust devil lifting provides a near constant background dust opacity throughout the Martian year. When each of the dust devil lifting schemes is tuned to match observed northern summer globally and zonally averaged opacities, the simulated opacities during southern spring and summer are low compared to observations (Figure 10).

Table 2. Low Thermal Inertia Regions
SimulationTharsisArabiaElysium
KMHBASE0225INDEP−0.39−0.360.06
KMH01INDEP−2.08−2.86−0.96
KMH035INDEP−1.94−1.87−1.07
KMH05INDEP−2.0−1.64−0.92
NEW01INDEP−2.48−0.52−0.15
NEW0225INDEP−0.05−0.070.39
NEW035INDEP−0.16−0.300.03
NEW050INDEP−0.77−0.43−0.02
KMH0225DUSTY0.32−1.78−0.87
KMH0225NODEV0.820.500.51

[49] The threshold-independent dust devil lifting parameterization allows dust to be lifted from a location whenever the surface heat flux is positive (upward heat transfer from the surface to the atmosphere) and the depth of the PBL is nonzero. Therefore dust devil lifting occurs throughout the day. The simulated diurnal peak in the dust devil lifting rate at the grid point nearest the Pathfinder site during the season of the Mars Pathfinder (MPF) mission occurs between 12 pm and 1 pm local time, which is consistent with the observed peak of dust devil occurrence (Figure 11).

Figure 11.

Pathfinder diurnal dust devil lifting rate. Solid curve indicates dust devil lifting rate (kg m−2 s−1) as a function of local time at the model grid point nearest the Mars Pathfinder landing site for a simulation with dust devil lifting only. Histogram indicates observed diurnal occurrence of dust devils during the Mars Pathfinder mission [Murphy and Nelli, 2002].

[50] The simulated spatial pattern and seasonal variation of dust devil lifting with the independent scheme compares well to observed dust devils and dust devil tracks [Fisher et al., 2005]. Of the nine regions included in Fisher et al.'s dust devil survey, the region with the greatest number of observed dust devils is Amazonis (25–45°N; 145–165°W). This region of peak dust devil activity is well reproduced by the model (Figure 12). Solis is a region of low dust devil activity in both Fisher et al.'s [2005] survey and in the threshold-independent dust devil lifting simulations presented here (Figure 13). Observations show that dust devils are most likely to exist in a region during its local summer [Fisher et al., 2005]; this seasonal variation is seen in the model results for both the individual regions used in Fisher et al.'s survey and hemispherically (Figures 13 and 14). Simulated dust devil activity at the Spirit landing site increases at Ls = 180, which corresponds to the season during which Spirit began observing dust devils. The simulated seasonality of dust devils at the location of Opportunity (in Meridiani) is suppressed relative to Spirit's location due to its close proximity to the equator. The lack of simulated dust devil seasonal variation at the grid point nearest Opportunity is consistent with the lack of observed dust devils by Opportunity. The simulated annual mean dust devil lifting rate in the northern hemisphere is a factor of 1.3 greater than the simulated annual mean dust devil lifting rate in the southern hemisphere. The simulated southern hemisphere dust devil lifting rate maximum occurs at Ls = 240 and is 74% above the annual simulated mean southern hemisphere dust devil lifting rate. The simulated northern hemisphere dust devil lifting rate maximum occurs at Ls = 70 and is 18% above the annual simulated mean northern hemisphere dust devil lifting rate. These simulated results suggest that there is more relative seasonality of dust devil lifting in the southern hemisphere but more overall dust lifted by dust devils in the northern hemisphere.

Figure 12.

Depth of dust removed (μm) by dust devils during each of eight Martian seasons for the threshold-independent dust devil lifting simulation. Each curve represents one of the regions included in the Fisher et al. [2005] survey. Numbers in the legend correspond to the labeled regions in Figure 13. Solid curves are northern hemisphere locations; dashed curves are southern hemisphere locations.

Figure 13.

Map of the depth of dust removed annually by dust devils for the threshold-independent dust devil lifting simulation. The heavy black boxes outline regions used in the Fisher et al. [2005] survey and Figure 12.

Figure 14.

Hemispheric dust devil lifting rate (kg m−2 s−1) as a function of Ls for a threshold-independent dust devil lifting simulation (solid curves) and a threshold-dependent dust devil lifting simulation (dot-dashed curve).

[51] The threshold-dependent dust devil parameterization of Newman et al. [2002] differs from the threshold-independent dust devil parameterization in that it requires that a tangential wind speed be exceeded before lifting occurs. Both the tangential wind speed and the tangential wind speed threshold are predicted by the model (see section 2.3.4.2). Implementing this scheme yields results that are similar in some respects to those produced by the threshold independent scheme. However, there are also substantial differences between the results produced by the two schemes. The threshold dependent scheme predicts that dust devil activity is highly concentrated in the northern hemisphere (specifically in the low thermal inertia continents of Tharsis, Arabia, and Elysium) (Figure 15). The maximum dust devil lifting occurs in Amazonis (Figure 15), which is consistent with observations [Fisher et al., 2005] and also with results from the threshold-independent dust devil simulation. In further agreement with the Fisher et al. [2005] survey, there is a seasonal dependence on the dust devil activity in Amazonis and the other northern hemisphere locations included in the survey. However, the dust devil activity in the southern hemisphere is muted in comparison to the northern hemisphere. Regions in the southern hemisphere that show seasonality in the Fisher et al. [2005] survey do not show a seasonal dependence in this simulation (Figure 16). The hemispheric seasonality that was cleanly predicted by the threshold independent dust devil lifting scheme is not reproduced by the threshold-dependent scheme (Figure 14). On the basis of the lack of seasonality predicted by this dust devil scheme into this model, the threshold-dependent scheme was not used in further simulations. The threshold-independent scheme straightforwardly depends on the vertical sensible heat flux and the depth of the boundary layer and exhibits good agreement with available dust devil observations. Therefore the threshold-independent scheme is employed in all simulations that include dust devil lifting.

Figure 15.

Map of the depth of dust removed annually by dust devils for the threshold-dependent dust devil lifting simulation. The heavy black boxes outline regions used in the Fisher et al. [2005] survey and Figure 16.

Figure 16.

Depth of dust removed (μm) by dust devils during each of eight Martian seasons for the threshold-dependent dust devil lifting simulation. Each curve represents one of the regions included in the Fisher et al. [2005] survey. Numbers in the legend correspond to the labeled regions in Figure 15. Solid curves are northern hemisphere locations; dashed curves are southern hemisphere locations.

4.2. Wind Stress Threshold Parameterizations

[52] Observations of dust lifting events provide constraints on the locations of wind stress dust lifting [Cantor et al., 2001]. Since the wind stress threshold controls the spatial extent of Mars' surface that can participate in wind stress lifting (i.e., higher threshold decreases the spatial extent of the simulated wind stress lifting), these observational constraints can help dictate the choice of wind stress lifting threshold. A wide range of stress threshold parameters have been used in previous dust cycle studies. Newman et al. [2002, 2005] use τ* = 10 mN m−2 and Basu et al. [2004] used a range of values between 30 and 60 mN m−2. The use of a stress threshold of 22.5 mN m−2 in previous Ames GCM studies and the baseline simulation presented here was based on physical arguments anchored in experimental results [Greeley and Iversen, 1985; Haberle et al., 2003; Murphy, 1999]. Here, we show that simulations using this threshold yield spatial lifting pattern results that are in good agreement with observations during the season Ls = 100–280 [Cantor et al., 2001].

4.2.1. KMH Lifting Scheme

[53] Several wind stress thresholds for dust lifting were explored using the KMH lifting scheme within the context of this fully interactive dust cycle model including dust devil lifting. In all cases, wind stress lifting is required to simulate the observed peak in opacity during southern summer. Stress thresholds ranging from τ* = 10 to 50 mN m−2 were used; the efficiency factor αW was tuned in each simulation to produce a dust cycle consistent with a nondust storm year as observed by TES (Figure 17; see Table 1). All stress threshold simulations discussed here included the threshold-independent dust devil scheme.

Figure 17.

Globally averaged 9 μm opacity for the (top) KMH and (bottom) Newman lifting schemes. Wind stress thresholds of 10, 22.5, 35, and 50 mN m−2 are represented by dotted, solid, dashed, and dot-dashed lines, respectively. The peak globally averaged opacity for the 50 mN m−2 threshold simulation with the Newman lifting scheme is approximately 0.75 and occurs at Ls = 285.

[54] The peak of the simulated globally averaged 9 μm opacity occurs near Ls = 290 in all of the tuned threshold simulations, which is approximately 30 degrees of Ls later than the observed peak [Smith, 2004]. This peak is produced by wind stress lifting in all simulations. In the simulations with higher stress thresholds (i.e., >22.5 mN m−2), the opacity is maintained by dust devil lifting during northern spring and summer. In the 10 mN m−2 simulation, additional dust lifting occurs during these seasons due to wind stress lifting. However, the opacity during these seasons is still too low compared to observations without dust devil lifting.

[55] As the stress threshold increases, the fraction of the planet's surface area that participates in wind stress lifting decreases (Figure 18). When the pattern of wind stress lifting for all threshold simulations is compared to Cantor et al.'s [2001] catalog of wind lifting dust events during the season Ls = 100–280, the 22.5 mN m−2 threshold simulation reproduces the observations better than the other threshold simulations. With a stress threshold of 10 mN m−2, wind stress lifting occurs in nearly all locations on the surface, which is contrary to Cantor et al.'s observed lack of dust lifting events in the equatorial regions. With stress thresholds of 35 and 50 mN m−2, wind stress lifting only occurs in limited longitudinal regions in both the northern and southern midlatitude to high-latitude regions. Observations of dust events are much more longitudinally extensive at these latitudes. On the basis of these results, the 22.5 mN m−2 stress threshold simulation was chosen to be the simulation used for extensive analysis.

Figure 18.

Maps of deflation due to wind stress (Ls = 100–280; contours and shading) and MOC observations of dust activity during this season (asterisks) for wind stress thresholds of 22.5, 10, 35, and 50 mN m−2 (top to bottom) with the KMH wind stress lifting scheme.

[56] In order to investigate nonlinear effects between the wind stress threshold and efficiency factor on the spatial pattern of wind stress lifting, a series of simulations was conducted employing each of the stress threshold values discussed above with a single efficiency value of 0.1. With a constant efficiency factor, the atmospheric dust load increased as the wind stress threshold decreased. While the magnitude of wind stress lifting for each threshold case changed depending on the efficiency factor used, the spatial pattern of wind stress lifting did not. Therefore to first order the spatial pattern of wind stress lifting depends only on the wind stress threshold and not on the efficiency factor.

4.2.2. Newman Lifting Scheme

[57] Simulations were conducted using the Newman wind stress lifting scheme with the same four stress thresholds that were used with the KMH scheme: 10, 22.5, 35, and 50 mN m−2. Since Newman et al. [2005] used a stress threshold of 10 mN m−2, their efficiency factor of 2.5 × 10−6 m−1 was employed in the 10 mN m−2 simulation presented here. The simulated dust cycle with this pair of parameters produced a reasonable dust cycle (Figure 17). Compared with the KMH baseline simulation, this simulation produces a more continuous increase in globally averaged opacity during southern spring but peaks at approximately the same Ls. The 22.5, 35, and 50 mN m−2 stress threshold simulations were tuned to produce a dust cycle similar to a nondust storm year as observed by TES by increasing the efficiency factor (Figure 17; see Table 1). All simulations included the threshold independent dust devil lifting scheme (with the same efficiency factor) that was used in the baseline simulation.

[58] The simulated dust cycles produced by the Newman lifting scheme with stress thresholds of τ* = 10, 22.5, 35, and 50 mN m−2, are very similar to those produced by the KMH lifting scheme with the same stress thresholds (but with different efficiency factors). The seasonal peak of the globally averaged opacity in simulations using the Newman lifting scheme peak at approximately the same Ls as the simulations using the KMH lifting scheme (Figure 17). In all of the τ* = 22.5, 35, and 50 mN m−2 simulations (i.e., with the KMH and Newman schemes), the northern spring and summer background opacity is maintained by dust devil lifting while the annual peak opacity is largely due to wind stress lifting in northern autumn and winter. In both τ* = 10 mN m−2 simulations (i.e., with the KMH and Newman schemes), the background opacity during northern spring and summer has a contribution from wind stress lifting but is primarily maintained by dust devil lifting. The peak opacity in northern autumn and winter is almost entirely due to wind stress lifting.

[59] The spatial pattern of preferred wind stress lifting regions is nearly identical with both lifting schemes when the same stress threshold is employed. This suggests that the spatial pattern of wind stress lifting depends on the stress threshold employed and that secondary nonlinear effects are not important. The fraction of the planet's surface area that participates in wind stress lifting increases as the stress threshold decreases (Figure 19). The τ* = 10 mN m−2 simulation with the Newman scheme (like with the KMH scheme) produces wind stress lifting over almost all of the planet during the season Ls = 100–280, which is contrary to observations [Cantor et al., 2001]. The fraction of the planet that participates in wind stress lifting in the τ* = 35 and 50 mN m−2 simulations is too small to match observations. For this reason, we argue that τ* = 22.5 mN m−2 is the best choice for a stress threshold in this model regardless of the lifting scheme employed. While this stress threshold reproduces the observed spatial extent of wind stress lifting, we note once again that the interannual variability of global dust storms is not simulated.

Figure 19.

Maps of deflation due to wind stress (Ls = 100–280; contours and shading) and MOC observations of dust activity during this season (asterisks) for wind stress thresholds of 22.5, 10, 35, and 50 mN m−2 (top to bottom) with the Newman wind stress lifting scheme.

[60] The Newman and KMH lifting schemes produce similar simulated dust cycles. Since we have a more detailed understanding of KMH scheme's sensitivities, we chose to use it over the Newman scheme.

5. Sensitivity Studies of Simulated Surface Dust Reservoirs

[61] Multiple simulations with a wide range of lifting parameters (i.e., wind stress threshold, efficiency factor, etc.) were conducted in order to test the robustness of the pattern of net surface dust deflation and deposition presented in section 3. In particular, we focused on the simulated net dust deflation in the three low thermal inertia continents of Arabia, Tharsis, and Elysium. The simulated deflation/deposition rates for these regions are summarized in Table 2.

[62] The effects of varying the wind stress threshold (using both the KMH and Newman lifting schemes) on the pattern of net deposition and deflation are shown in Figures 20 and 21. Locations that remain deposition regions in all stress threshold simulations (all include dust devil lifting) include the center of Hellas and the north and south polar regions. In all simulations, large regions in and around the low thermal inertia continents are net deflation regions. When the wind stress threshold is large (i.e., 35 and 50 mN m−2), the majority of the equatorial to midlatitude regions have small amounts of simulated net deflation even though these regions do not experience wind stress lifting. In particular, all three low thermal inertia continents are net deflation regions (Figure 22). Dust devil lifting is responsible for keeping these locations from becoming deposition regions. In the small wind stress threshold simulation (i.e., 10 mN m−2), wind stress lifting is active nearly planet-wide; thus wind stress and dust devil lifting are responsible for deflating large sections of the low thermal inertia continents. The dust removed from these regions accumulates in the polar regions.

Figure 20.

Map of net annual dust deflation (shaded regions) and deposition (nonshaded regions) for the KMH wind stress lifting scheme simulations. The heavy black boxes outline the three low thermal inertia regions: Tharsis (left), Arabia (middle), and Elysium (right).

Figure 21.

Map of net annual dust deflation (shaded regions) and deposition (nonshaded regions) for the Newman wind stress dust lifting scheme simulations. The heavy black boxes outline the three low thermal inertia regions: Tharsis (left), Arabia (middle), and Elysium (right).

Figure 22.

Net surface dust depth deflated (negative) or deposited (positive) as a function of Ls in the three low thermal inertia regions for the four (10, 22.5, 35, and 50 mN m−3) KMH wind stress threshold simulations.

[63] Dust devil lifting is almost completely insensitive to the wind stress threshold in the simulations that were tuned to reproduce a nondust storm year. The total quantity of dust devil lifting remains approximately constant in all simulations. The spatial pattern of dust devil lifting is essentially unchanged in all stress threshold simulations.

[64] In addition to examining the effects of varying the wind stress threshold, the wind stress efficiency factor was increased by 50% in a simulation with a wind stress threshold of 22.5 N m–2. This configuration was designed to produce a simulated dust cycle that more closely matched a global dust storm year (Figures 23 and 24). Although the magnitude of the simulated atmospheric dust opacity is considerably larger at all latitudes during southern summer, the peak simulated zonally averaged opacities occur at approximately the same Ls as during a simulated nondust storm year. Dust devil lifting is slightly suppressed when the local atmospheric dust load is large (i.e., during southern spring and summer) due to the thermal stability of the near surface atmosphere. Wind stress lifting remains very small during northern spring and summer, and the dust devil lifting remains nearly the same as the baseline case during these seasons. While some differences are seen in the spatial pattern of net deflation and deposition, the low thermal inertia regions remain largely deflation regions in this high efficiency factor “dust storm” simulation (Figure 20).

Figure 23.

Zonally averaged 9 μm opacity for the high-efficiency τ* = 22.5 mN m−2 baseline simulation. Opacities have been scaled to the 6.1 mbar pressure level.

Figure 24.

Globally averaged 9 μm opacity for the high-efficiency τ* = 22.5 mN m−2 baseline simulation. Opacities have been scaled to the 6.1 mbar pressure level.

6. Discussion

6.1. Dust Cycle

[65] The simulated dust cycle in the baseline simulation agrees reasonably well with available observations. Model results and observations exhibit a background atmospheric dust haze during northern spring and summer and a peak in the atmospheric dust load during southern spring and summer (Figures 2 and 3) [Smith, 2004]. The spatial pattern of simulated wind stress lifting during Ls = 100–280 compares well to a catalog of observed dust lifting events (Figure 18) [Cantor et al., 2001] during the same time of year. Additionally, the spatial pattern and seasonality of dust devil lifting, as simulated with the threshold-independent Newman et al. [2002] scheme, agrees favorably with a recent survey of dust devils and dust devil tracks in MOC images (Figures 12 and 13) [Fisher et al., 2005]. Simulated temporally and zonally averaged temperatures at the equinoxes and solstices reflect the corresponding TES-derived temperature fields (Figure 4) [Banfield et al., 2004].

[66] While broad features of the dust cycle appear to be well captured by the Ames GCM, there are also model shortcomings that deserve discussion. The simulations presented here do not exhibit substantial year-to-year variations in the dust cycle, which is inconsistent with observations. In all simulations, the simulated peak in the zonally and globally averaged opacity occurs tens of degrees of Ls later than TES-derived zonally and globally averaged opacity peak [Smith, 2004] for Mars year 24 and tens of degrees of Ls earlier than TES-derived zonally and globally averaged opacity peak for Mars year 26 (Figure 2). The seasonal timing of peak opacity is independent of the wind stress lifting scheme (i.e., KMH or Newman) or wind stress threshold chosen (Figure 17).

[67] Although the model-predicted spatial pattern of wind stress lifting generally agrees with observations [Cantor et al., 2001] during the Ls range of the observations, there are also localized regions of observed lifting events that are not well represented by the model. Lifting at high latitudes is underpredicted by the model. Since the CO2 ice polar caps are not well represented at the horizontal resolution used in these GCM simulations (5° in latitude × 6° in longitude), increasing the model's resolution may increase cap edge lifting. Lifting is also not predicted to be as longitudinally extensive throughout the southern midlatitude band as suggested by observations [Cantor et al., 2001]. These regions are also trouble spots for other GCM dust cycle studies [Basu et al., 2004; Newman et al., 2005].

[68] These discrepancies between the simulated and observed dust cycle could indicate that there are critical physical processes that are not included. There is a wide range of feedbacks between the dust cycle and other parameters (e.g., albedo, surface roughness, the water cycle, etc) that have not yet been studied in detail. The model includes a spatially and temporally uniform surface roughness due to the uncertainties involved in deriving the “true” Martian surface roughness spatial pattern. Surface roughness affects the model-calculated surface stress; thus a spatially varying surface roughness field could affect the simulated dust cycle. The spatial pattern of surface albedo has been observed to change on subannual timescales [Lee, 1987; Smith, 2004; Kahre et al., 2005a] due to the redistribution of surface dust. These albedo changes affect the thermal state of the atmosphere and the depth of the planetary boundary layer [Kahre et al., 2005a]. Therefore feedbacks between dust deposition and albedo could potentially alter the simulated dust cycle. Interactions between the dust and water cycles through clouds could also prove to be an important model addition.

6.2. Surface Dust Reservoirs

[69] A robust result of all simulations presented here is that, on average, the three low thermal inertia regions are not accumulating dust. In fact, these simulations indicate that these regions might be experiencing small magnitude net deflation (Figures 20 and 21). The pattern of net deflation/deposition of the baseline simulation can be directly compared to Figure 18 of Basu et al. [2004]. While many regions of net deflation agree between the two simulated results, there are also regions of disagreement. Regions of net deflation in our baseline simulation that are in good agreement with Basu et al.'s [2004] results include the southwest and northeast interior slopes of Hellas, the southwest interior slope of Argyre, the slopes of Tharsis, and regions in Elysium. Regions of net deposition that are in good agreement include the north and south polar regions and the center of Hellas. Chryse is a net deflation region due to strong winds in the GFDL model [Basu et al., 2004]. However, the Ames GCM does not predict particularly large amounts of wind stress or dust devil lifting at that location. Thus Chryse is a net deposition region in our baseline simulation. It is difficult to tell whether the three low thermal inertia regions are net deflation or net deposition on average in Basu et al.'s work, but as those authors note, their simulated net deflation/deposition spatial pattern does not correspond well to the locations of known surface dust deposits (i.e., the low thermal inertia regions). Therefore the result presented here that the low thermal inertia regions are not accumulating at the present time is broadly consistent with Basu et al.'s [2004] results.

[70] Simulated dust deposition rates can be directly compared to accumulation rates experienced by Martian landers. In the mission's first 36 sols (Ls = 140–160), the Mars Pathfinder Material Adherence Experiment (MAE) experienced a dust accumulation rate of 0.29% per sol (where sol is Martian solar day = 24.6 hours) [Landis and Jenkins, 2000]. This accumulation rate corresponds to an optical depth accumulation rate of 0.004 τ sol−1. The simulated dust deposition rate at the location of MPF in the baseline simulation during this time of year averaged over 36 sols yields a dust optical depth deposition rate of 0.003 τ sol−1. Both Spirit and Opportunity experienced a dust accumulation rate of approximately 0.2% per sol for the first 90 sols of operation [Arvidson et al., 2004a, 2004b], which correspond to a dust optical depth accumulation rate of 0.0028 τ sol−1. The simulated dust deposition rate at these locations during this time of year in our baseline simulation produces a dust optical depth deposition rate of0.0025 τ sol−1. On the basis of the assumptions stated here, the simulated dust deposition rates at all three landing sites are consistent with the observed dust deposition.

[71] The baseline simulation generates significant dust deposition within the north and south polar regions. The dust deposition rate remains nearly constant throughout the season of cap mass accumulation in both the northern and the southern hemispheres (Figure 8). Therefore the dust-to-ice ratio of the seasonal cap will be at a maximum early in the cap growth process when the rate of cap growth is at a minimum. Differing dust-to-ice ratios in the seasonal cap during its growth could result in time varying cap albedos during cap retreat. These temporal albedo changes could alter the cap sublimation rate [Hourdin et al., 1995]. Implementing such an albedo feedback into a GCM would enable more detailed comparisons between observed and simulated cap retreat rates.

[72] The total simulated north polar dust deposition during one year is 7.75 × 10−3 kg m−2 yr−1. This value is significantly smaller than the value of 0.2 kg m−2 yr−1 suggested by Pollack et al. [1979]. Their deposition rate was based upon observations obtained over the course of the two global dust storms that occurred during the first Viking year and the assumption that all of the suspended dust was deposited in the north polar region. The simulations presented here show that this is not a good assumption; only 2% of the deposition during one year of the baseline simulation is in the north polar region. Polar deposition predicted by the high efficiency simulation that was designed to produce a global dust storm increases by a factor of 1.2 over the baseline simulation, which is still significantly lower than Pollack et al.'s [1979] prediction. This difference between our GCM results and the Pollack et al. [1979] prediction indicates that the GCM simulated net transport of dust from the southern hemisphere to the northern hemisphere is not as efficient as Pollack et al. [1979] assumed. Within a fraction of a percent, the baseline simulation presented here does not indicate that there is an annual net interhemispheric transport of dust.

6.3. Importance of Dust Devils and the Stability of Low Thermal Inertia Regions

[73] Dust devils play a critical role in the simulations presented in this paper. In agreement with Basu et al. [2004], simulation results indicate that dust devils are responsible for maintaining the background opacity during northern spring and summer. Simulated dust devil lifting contributes half of the total dust lifted during an annual cycle during which no global dust storm occurred. These simulations also suggest that dust devils play an important role in the present-day pattern of net surface dust deflation and deposition. To investigate this further, we conducted a simulation with wind stress lifting only using the same stress threshold and efficiency factor as the baseline simulation. Without dust devil lifting, the low thermal inertia regions become deposition regions as predicted by Haberle et al. [2003] (Figure 25). Therefore dust devil lifting is responsible for the slight deflation rates in three low thermal inertia continents in the baseline simulation.

Figure 25.

Map of net annual dust deflation (shaded regions) and deposition (nonshaded regions) for a simulation without dust devil lifting. The heavy black boxes outline the three low thermal inertia regions: Tharsis (left), Arabia (middle), and Elysium (right).

[74] Christensen [1986] proposed that the low thermal inertia continents are undergoing alternating periods of dust deposition and removal. If this cyclic process corresponds to orbital variations, some combination of orbital parameters must produce deflation in these regions. At high obliquity, it has been shown that wind stress lifting dominates over dust devil lifting globally [Newman et al., 2005]. Additionally, it has been shown that wind stress lifting is not substantial in these low thermal inertia regions at any obliquity [Haberle et al., 2003; Newman et al., 2005]. Therefore it may be that periods of low obliquity are responsible for the periods of deflation. According to the results presented here, the current orbital configuration might be near the transition point between these regions being deposition or deflation regions.

6.4. Closing the Martian Dust Cycle?

[75] The simulations presented here predict the existence of annual net surface dust sources and sinks, which strongly suggests that the Martian dust cycle is not closed regionally on annual timescales. However, model results do suggest that the dust cycle is closed hemispherically on an annual basis. On the basis of these results, one may ask what timescale is required for regional closure of different spatial scales. This question is made more complicated by the likelihood that each location or region on the planet is closed on a different timescale and by the possibility that some locations are closed while others are not. For instance, we have postulated that the depth evolution of the three low thermal inertia continents is controlled by changes in orbital obliquity. Thus these regions may be closed on the timescales associated with changes in orbital obliquity. However, climate change associated with orbital oscillations may not close all locations on the planet. For example, model predicted preferred wind stress lifting regions (e.g., the flanks of Hellas) may not become dust accumulation regions under any orbital configuration. Other factors that occur on long timescales such as atmospheric evolution, the creation or removal of mobile dust particles, and interactions with the water cycle will also likely play important roles.

[76] In addition to processes that occur over long timescales, there are processes occurring on annual or multiannual timescales that are not currently included in the GCM. Albedo and thermal inertia feedbacks have thermodynamic effects on the local atmosphere [Smith, 2004; Kahre et al., 2005a] and could affect the simulated dust cycle and the simulated spatial pattern of annual net dust deflation/deposition. Additionally, including a spatially and/or temporally varying surface roughness field and wind stress threshold field could affect the results presented here. Including these processes in future GCM studies will help illuminate their relative importance on the Martian dust cycle and the evolution of surface dust reservoirs.

7. Conclusions

[77] The Martian dust cycle has been investigated with a fully interactive (i.e., the lifting, transport, and sedimentation of radiatively active dust) version of the NASA Ames general circulation model. Two dust lifting mechanisms, dust devil and wind stress lifting, have been implemented and the effects of each mechanism on the resulting simulated dust cycle have been studied in detail. Particular focus has been given to understanding the effects of both wind stress and dust devil lifting on the simulated pattern of net surface dust deposition and deflation.

[78] The main results of this work are the following:

[79] 1. Both wind stress and dust devil lifting are required to accurately simulate the zonally and globally averaged spatial and temporal pattern of atmospheric dust opacity. Dust devil lifting maintains the background haze of dust during northern spring and summer, and wind stress lifting produces the simulated peak annual opacity during southern spring and summer (Figure 3). The simulated contributions to the total dust lifted annually from dust devils and wind stress lifting are approximately equal.

[80] 2. The KMH and Newman wind stress lifting schemes yield similar simulated dust cycles when the same wind stress thresholds are factors are implemented and the efficiency tuned to match globally averaged opacities that are consistent with a nondust storm year as observed by TES (Figure 17).

[81] 3. Using a wind stress dust lifting threshold of 22.5 mN m−2 yields a pattern of surface dust deflation that reflects observed locations of wind stress lifting events during the season Ls = 100–280 [Cantor et al., 2001] more closely than using higher or lower wind stress thresholds (Figures 18 and 19).

[82] 4. The threshold-independent dust devil lifting scheme produces spatial and temporal (daily and seasonal) patterns of dust devils that are consistent with observations [Murphy and Nelli, 2002; Fisher et al., 2005]. With this scheme, the peak of dust devil activity occurs during local summer, and Amazonis is a particular active region for dust devils (Figures 12 and 13). Both of these features are in agreement with observations [Fisher et al., 2005]. Additionally, the simulated peak of the diurnal dust devil lifting rate at the Mars Pathfinder site occurs at the same time as the observed peak in dust devil activity during the MPF mission (Figure 11) [Murphy and Nelli, 2002].

[83] 5. The three low thermal inertia continents (Arabia, Tharsis, and Elysium) on average experience no net accumulation of dust when dust devil lifting is included. This is a robust result; sensitivity studies with both the KMH and Newman wind stress lifting schemes indicate that these regions do not accumulate dust for any reasonable combination of wind stress threshold and wind stress efficiency factor (Figures 20 and 21). The lack of net accumulation in these regions is due to dust devil lifting. When dust devil lifting is not included, all three of these regions become deposition regions (Figure 25).

Acknowledgments

[84] We would like to thank Phil Christensen and Ron Greeley for helpful discussions regarding this work. Funding was received from the NASA Graduate Student Researchers Program (103854/NAG5-50457) and NASA's Planetary Atmospheres Program (103885/NAG5-12123). We thank Scot Rafkin and an anonymous reviewer for their helpful comments and suggestions regarding the submitted manuscript.

Ancillary