Thermophysical Analysis of Gale Crater Using Observations from
TES, THEMIS and the Mars Science Laboratory Curiosity Rover
Edward M. Barratt BSc (Hons), University of Edinburgh, 2011
December 5, 2013
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirement for the degree of
Master of Science
Department of Astrophysics and Planetary Science
2013 This thesis entitled
Thermophysical Analysis of Gale Crater Using Observations from
TES, THEMIS and the Mars Science Laboratory Curiosity Rover
written by Edward Mark Barratt
has been approved for the Department of Astrophysical and Planetary Science
Prof. Brian Hynek
Signature Date
Dr. Nathaniel Putzig
Signature Date
Prof. Shijie Zhong
Signature Date
Dr. Mike Mellon
Signature Date
The final copy of this thesis has been examined by the signatories, and we
find that both the content and the form meet acceptable presentation standards
of scholarly work in the above mentioned discipline.
ii Edward Mark Barratt (Masters of Science, Department of Astrophysics and Planetary Science)
Thermophysical Analysis of Gale Crater Using Observations from
TES, THEMIS and the Mars Science Laboratory Curiosity Rover
Thesis directed by Professor Brian Hynek of the department of Geological Sciences.
Abstract I created a web-based interface to the MARSTHERM one-dimensional numerical thermal model, which cal- culates surface and atmospheric temperatures of Mars for a set of user-specified conditions. The website also provides access to tools which I adapted to allow users to automatically derive thermal inertia from images taken by the Mars Odyssey Thermal Emission Imaging System (THEMIS), and access to existing maps of thermal inertia derived from Mars Global Surveyor Thermal Emission Spectrometer (TES) observations. To demonstrate the capabilities of the tools provided on the website, I conducted a case study investigating the thermal inertia within Gale Crater, using observations from TES, THEMIS, and the Mars Science Lab- oratory (MSL) Ground Temperature Sensor (GTS). Seasonal variations in TES-derived thermal inertia in the vicinity of the MSL landing site are consistent with a layer of sand or dust over rock, an interpretation supported by diurnal temperature variations recorded by GTS on MSL sol 30. However, diurnal tempera- ture variations from elsewhere along the MSL traverse route could not be modeled by simple two-component structures. Seasonal variations in TES-derived thermal inertia support the hypothesis that the thickness of a surface layer of dust likely increases with elevation on Mount Sharp.
iii Acknowledgments None of this work would have been accomplished without the advice, patience and assistance of Than
Putzig at the Southwest Research Institute, who generously entrusted me with the construction of the MARS-
THERM website whilst supporting and encouraging my research efforts. I am also grateful for the guidance and hard work of Brian Hynek and Fran Bagenal, who both assisted me in navigating the complications of educational bureaucracy and were quick to offer advice when I needed it. Thanks are also due to the other members of my committee, Mike Mellon and Shijie Zhong. And also to my good friends and housemates at the Mackshack for making the last two years in Boulder so enjoyable.
iv Contents
List of Figures vii
List of Tables viii
1 Introduction 1
2 The MARSTHERM Thermal Model 5
2.1 Thermal Inertia Derivation ...... 7
2.1.1 Single Point Thermal Inertia Derivation ...... 7
2.2 Surface Heterogeneity ...... 8
2.3 Instruments ...... 11
2.3.1 TES ...... 11
2.3.2 THEMIS ...... 11
2.3.3 REMS ...... 12
2.3.4 MOLA ...... 13
2.3.5 Imaging Instruments ...... 13
3 The MARSTHERM Website 14
3.1 Project Designation ...... 16
3.1.1 Project Maps ...... 16
3.1.2 Dust Opacity ...... 16
3.1.3 Two Component Modeling ...... 17
3.2 THEMIS Image Processing ...... 17
3.2.1 Download and Pre-Processing ...... 17
3.2.2 Thermal Inertia Derivation ...... 19
3.2.3 Output Files ...... 20
3.2.4 Known Issues ...... 21
4 Gale Crater 26
4.1 Mount Sharp ...... 26
4.1.1 Thermal Inertia of Mount Sharp ...... 28
4.2 Northern Crater Floor ...... 29
v 4.2.1 Peace Vallis and Peace Vallis Fan ...... 32
4.3 Diurnal Temperature Cycles ...... 42
4.3.1 MSL sol 30 ...... 43
4.3.2 MSL sol 100 ...... 46
5 Conclusions 50
6 References 52
vi List of Figures
1 Seasonal atmospheric pressure within Gale Crater...... 3
2 Thermal inertia and particle-size scaling...... 4
3 Model diurnal surface temperatures at lookup-table nodes...... 9
4 THEMIS image times and seasons for Gale Crater...... 12
5 MARSTHERM website and model temperatures in Gale Crater...... 15
6 TES dust opacity history within Gale Crater...... 19
7 THEMIS thermal inertia products for Gale Crater from the MARSTHERM website...... 22
8 Histogram of global TES-derived thermal inertia values...... 23
9 Simulated Histogram showing Shark-fin Artifacts...... 25
10 Gale Crater THEMIS visible mosaic...... 27
11 THEMIS brightness temperature and derived thermal inertia for Gale Crater...... 30
12 THEMIS brightness temperature and derived thermal inertia for a profile across Mount Sharp. 30
13 TES-derived thermal inertia of Mount Sharp...... 31
14 Seasonal variations in apparent thermal inertia on Mount Sharp...... 31
15 Peace Vallis thermal inertia...... 33
16 Peace Vallis elevation from HRSC DTM...... 34
17 Histograms of Peace Vallis thermal inertia...... 35
18 The high thermal inertia area of the Peace Vallis fan...... 36
19 Histograms of thermal inertia in the Peace Vallis fan...... 37
20 HiRISE images of the high thermal inertia area of the Peace Vallis fan...... 38
21 Annual-median TES-derived thermal inertia on the Peace Vallis fan...... 39
22 Seasonal variations in apparent thermal inertia on the Peace Vallis fan...... 41
23 HiRISE image of the Curiosity traverse...... 42
24 Mosaic of MSL sol 30 site...... 44
25 GTS diurnal temperature cycle at MSL sol 30 site...... 45
26 Left navigation camera image of MSL sol 100 site at Rocknest...... 47
27 GTS diurnal temperature cycle at MSL sol 100 site, with single component model curves. . . 48
28 GTS diurnal temperature cycle at MSL sol 100 site, with two component model curves. . . . 49
vii List of Tables
1 Thermal model parameters...... 6
2 Thermal model lookup-table properties...... 8
3 Idealized surface materials...... 10
4 Apparent thermal inertia of Peace Vallis Fan...... 40
5 MSL diurnal temperature cycles...... 43
viii 1 Introduction
Since 1996, when Mars Global Surveyor (MGS) began the first successful mission to Mars since the Viking era, Mars has been the target of 9 successful missions. Two rovers are currently exploring the surface of the planet, and three orbiters are in place continually gathering and returning data. With still more missions planned by NASA, ESA, and the Russian and Indian space agencies, no other planet apart from our own has ever been subject to such intense and active exploration. These missions have revealed clear evidence that the past climate of Mars was very different from how it is now. With newer data the sparse valley networks observed by the Viking orbiters (e.g. Masursky et al. 1977) have been revealed as integrated and mature drainage systems indicative of flowing surface water (Hynek et al. 2010). Spectral data indicates that clay minerals, which only form in aqueous conditions, occur in sedimentary rocks on the surface (e.g. Michalski et al. 2005; Milliken et al. 2010). The global distribution of deltas and valleys even suggests the possibility of an ancient Martian ocean (Di Achille & Hynek 2010). All of this evidence provides proof that Mars was at one time warmer and wetter than it is now. Understanding more about these early conditions can help us to theorize about the habitability of the early planet. Could Mars have once hosted life? And if so, are there signs that it did? Continued research on the geology of Mars can help to answer these questions, while answers to them will also help us to understand the nature of life on our own planet. This knowledge will provide us with a better idea of past or present conditions that may support life on other planets in our own solar system and beyond.
While rovers provide a means of investigating the geology of the surface in situ at a few small sites, investigations over broader regions are predominantly limited to using data from remote sensing instruments on the Martian orbiters or Earth based telescopes. Surface temperature is one measurement that can be derived from remote-sensing observations. The temperature of a planetary surface is controlled by a balance of the outward radiated heat from the surface with the inward heat flux due to solar insolation (SM ) and
− atmospheric radiation (FIR), while also accounting for heat exchange with the subsurface by conduction, and any other heat sources. In the case of Mars, another heat source that must be considered is latent heat due to seasonal CO2 condensation and sublimation. This balance provides an equation for the surface temperature (Ts) (e.g. Mellon et al. 2008):
r 4 − ∂m π ∂T σTs = SM (1 − A) cos i + FIR + L + I 0 (1) ∂t Γ ∂Z Z0=0 where is the surface emissivity, σ is the Stefan-Boltzmann constant, A is the albedo, i is the solar incidence
1 angle and L is the latent heat of CO2 sublimation or condensation, and m is the mass of CO2 frost. I is the thermal inertia, Γ is the seasonal (or diurnal) period, and T is the subsurface temperature while Z0 is depth normalized by the seasonal (or diurnal) thermal skin depth δ, where δ is given by:
r I Γ δ ≡ (2) ρc π
where ρ is the density and c is the heat capacity of the surface materials. For two frost-free regions in
close proximity, the first three terms on the right-hand side of Eq. 1 will be approximately equal, so that
temperature differences between the two regions will be controlled by thermal inertia. Thermal inertia is a
bulk property that controls how a volume of material stores and conducts heat. The surface temperature of
an ideal object with negligible thermal inertia would respond instantaneously to radiative forcing, depending
solely on incident radiation and the object’s albedo. Real materials have non-zero conductivity (k) that
allows heat to pass into their interiors, while their density and heat capacity allow them to store heat.
Together, these three properties comprise the thermal inertia:
I ≡ pkρc (3)
In this work I will use the SI derived unit of thermal inertia, tiu:
tiu ≡ Jm−2K−1s−1/2 (4)
For geologic materials under Martian surface conditions, thermal inertia generally increases with grain size (Mellon et al. 2008). Presley & Christensen (1997) used laboratory measurements to show that, under
Martian conditions, the thermal conductivity can be estimated by:
0.6 (−0.11 log (P/c2)) k = (c1P )d 10 (5)
where P is atmospheric pressure in Pa, and d is the grain diameter in µm, yielding k in Js−1m−1. The
constants c1 and c2 depends on other properties of the material, especially the bulk density. For a medium well-packed density analogous to loose sediments on a planetary surface (as described by Presley & Chris-
−5 7 tensen (1997)), c1 = 7.97 × 10 and c2 = 1.08 × 10 Pa. The seasonal pressure variations on Mars can be taken from the fit to the first year of Viking Lander 1
measurements (Tillman et al. 1993) and scaled for elevation using a scale height of 10.8 km. Figure 1 shows
2 the range of pressures that this method produces for Gale Crater, with the atmospheric pressures ranging from 4.36–5.76 mbar on the south crater rim (elevation 0.1 km) to 7.41–9.78 mbar at the crater’s low point
(elevation -4.5 km).
Figure 1: The seasonal atmospheric pressure in Gale Crater, as calculated using the Tillman et al. (1993) fit and scaled to elevations within the crater. The southern autumn equinox occurs at Solar Longitude 0◦.
Figure 2 shows a model scaling between particle size and thermal inertia within these pressure ranges, demonstrating that changes in pressure within the crater and through the year will have a modest effect on the surface thermal inertia, compared to changes due to material properties such as grain size itself. To derive Eq. 5, Presley & Christensen (1997) used spheres of glass in a CO2 gas atmosphere, so it should be relevant to the analysis of sand dunes, dust, and other dry unbonded agglomerates. In such materials under
Mars surface conditions, conduction between grains is dominated by the conductivity of the gas in the pore spaces (Mellon et al. 2008). Conductivity may be increased by physically bonding particles, thereby filling the pore space with a material of greater conductivity than gas. A duricrust of material bonded in this way was observed at the Viking Lander 1 site (Shorthill et al. 1976), and its effect on the thermal inertia of the soil was described by Kieffer (1976). Duricrust was also observed at the Pathfinder landing site (Moore et al. 1999). Putzig & Mellon (2007) showed that large regions of the low to mid latitudes of Mars had thermophysical responses that could be indicative of a layer of duricrust overlying dust. Observations of Gale
Crater from orbit suggest that large regions of the crater floor are characterized by duricrust (e.g. Fergason et al. 2012; Anderson & Bell 2010). The likelihood of duricrust layers and other forms of heterogeneity
3 means that the use of thermal inertia to infer particle size should be considered with caution.
Figure 2: Relation between thermal inertia and effective particle size for minimum and maximum expected pressures in Gale Crater, assuming ρ = 1600 kg m−3 and c = 625 J kg−1K−1. Derived using Eq. 5 (Presley & Christensen 1997). Grain-size designations are standard Wentworth scale. Adapted from Kieffer (2013)
In Section 2, I will describe the numerical model that calculates surface and atmospheric temperatures and the derivation algorithm that is used to derive thermal inertia from temperature measurements of Mars.
In Section 3, I will describe the structure and the capabilities of the MARSTHERM website that I have developed to provide access to these and other thermophysical analysis tools. In Section 4, I will discuss a case study where I have used these and other tools to analyze surface properties within Gale Crater.
4 2 The MARSTHERM Thermal Model
The MARSTHERM one-dimensional numerical thermal model was adapted by Putzig & Mellon (2007)
from a previous model developed by Mellon et al. (2000). It calculates surface and brightness temperature variations for seasonal and diurnal periods, under a defined set of conditions. Brightness temperature is surface temperature scaled by surface emissivity and attenuated by atmospheric opacity, and it is the temperature that one would calculate by measuring the infrared radiance from space and assuming a black body radiator. MARSTHERM calculates brightness and surface temperatures using user-specified values of surface properties (thermal inertia, albedo, emissivity, atmospheric pressure) and atmospheric dust opacity
(τD); water-ice-cloud opacity is not included in the current version of the model. Models can be run for a single latitude, or simultaneously for several regularly spaced latitudes. Longitude is not parameterized since outputs are for local solar times. Topography can be accounted for by adjusting slope azimuths and angles and surface pressure. The model can be accessed on the MARSTHERM website described in Section 3.
To calculate the temperature of a surface on Mars using Eq. 1 still poses some problems. It is possible to measure from orbit, or to make educated guesses about, the emissivity, albedo, thermal inertia and the frost covering for a particular time and place. The solar irradiation can be easily modeled from the eccentricity of the Martian orbit, while the solar incidence angle is a simple function of location, time of day and season.
But the subsurface temperature gradient must be calculated, and depends on the past variation of the temperature. The model must therefore be “spun up” over a few seasonal (or diurnal) cycles to eliminate the effects of initial conditions.
The atmospheric model incorporates solar and thermal-infrared radiative transfer in a dusty CO2 atmo- sphere (Pollack et al. 1990), as described by Mellon et al. (2000). Improvements to the near-surface and atmospheric portions of the model are being undertaken and once this work is completed, the revised model will replace the current one on the website.
Table 1 shows all of the parameters that may be specified for the thermal model, and those that are pre-programmed. In addition to physical properties, users may specify the number of times per day at which surface ground temperature is calculated, the number of iterations of the subsurface model per iteration of the atmospheric model, the interval in subsurface iterations at which diurnal temperatures are output, and the interval in days at which seasonal temperatures are output.
5 Table 1: Thermal Model Input Parameters. More details are on the website docu- mentation pages.
Name Units Range/Value User specified parameters Thermal Inertia (I) tiu 5–5000 Albedo of Bare Ground 0–1 Albedo of Bare CO2 Surface Ice 0–1 −2 Effective Semi-Infinite CO2 kg m 0–1000 Bare Ground Emissivity @ 15 µm 0–1 Bare Ground Emissivity elsewhere 0–1 Surface Slope Angle ◦ from horizontal 0–90 Surface Slope Azimuth ◦ clockwise from north 0–360 Surface Pressure (Γ) mbar 0–1000 Visible Dust Opacity (τD) 0–1000 Southern Most Latitude ◦ north from equator −90–90 Latitudes Per Run 1–37 Latitude Increment ◦ north 0–180 1 ◦ Solar Longitude (Ls) 0–360 Ground Time Steps per Day 1–10000 Ground Time Steps per Atmospheric...... Time Step 1–99 Seasonal Output Interval 2 sols 1–668 Diurnal Output Interval time steps 1–99 Pre-programmed parameters Soil Heat Capacity J kg−1 K−1 627.900 IR/Visible Dust Opacity @ 15µm 0.500 IR/Visible Dust Opacity Elsewhere 0.500 CO2 Ice Emissivity at 15µm 0.800 CO2 Ice Emissivity Elsewhere 0.800 Soil Density kg m−3 1.5×103 Initial Temperature K 200 Wind Speed m s−1 0.000 Seasonal Spin Up Time 2 Martian Years 3 Diurnal Spin Up Time 1 sols 10 1. Diurnal model only. 2. Seasonal model only.
6 2.1 Thermal Inertia Derivation
Early work to measure the thermal inertia of the Martian surface involved adjusting the albedo and thermal
inertia inputs for a thermal model similar to the one used here, in order to produce the best fit to acquired
temperature data for the surface (e.g. Palluconi & Kieffer 1981; Hayashi et al. 1995). This technique is only
feasible when temperature observations sample the same region frequently at multiple stages of a diurnal
cycle. The Viking Infrared Thermal Mapper data used by Palluconi & Kieffer (1981) and Hayashi et al.
(1995) was collected in such a way that this was possible (Kieffer 1976), but the MGS and the Mars Odyssey
spacecraft are is sun-synchronous orbits, which means that they sample regions away from the poles at nearly
the same local time of day on every pass, so that repeat observations of the same point occur at two times
of day at best and therefore the diurnal temperature variation is not well sampled by these spacecraft.
2.1.1 Single Point Thermal Inertia Derivation
Mellon et al. (2000) developed a method that allows thermal inertia to be derived from single-point temper-
atures for use with Mars Global Surveyor Thermal Emission Spectrometer (TES) observations. First, the
thermal model is used to compute temperature cycles for values across the expected range of albedo, thermal
inertia, surface pressure, dust opacity and latitude conditions. Model temperatures are stored for ranges of
time of day and season, producing a 7-dimensional lookup table. Each single-point temperature observation
is then correlated with modeled or measured values for each of the 6 dimensions apart from thermal inertia.
Thermal inertia is then interpolated from the lookup table for the specific set of parameters.
Derivation and mapping of thermal inertia by Mellon et al. (2000) and Putzig et al. (2005) found values
at the limits of the lookup table, which allowed values between 24 and 800 tiu. Among other updates to the
derivation algorithm, Putzig & Mellon (2007) extended the lookup table to allow thermal inertia between
5 and 5000 tiu, and used it to derive separate 2am and 2pm global maps of Martian thermal inertia at 20
pixels per degree (ppd) resolution. The current derivation code for both TES and THEMIS thermal inertia
uses the lookup table of Putzig & Mellon (2007). Table 2 provides details of the lookup table, including the
data source for interpolation along each dimension.
◦ For the latitude of Gale Crater and Ls = 0 , Figure 3 presents modeled surface temperatures at each node value for four dimensions of the lookup table, together with two panels showing the effect of surface slopes. The parameter with the largest effect on modeled surface temperatures is thermal inertia. A surface with lower thermal inertia will have colder nighttime temperatures and higher daytime temperatures. In this example, CO2 frost forms and limits the minimum temperature to ∼148 K at certain times of night for
7 Table 2: Thermal Model Lookup-table Properties
Dimension Nodes Range Data Source Mars Day 84 0–668 Spacecraft Ephemeris Mars Hour 24 1–24 Spacecraft Ephemeris Latitude 37 −90–90 Spacecraft Ephemeris Dust Opacity 3 0.0–0.8 TES data (Smith 2004) Pressure 5 0.8–15.0 Scaled from (Tillman et al. 1993) fit using a 20-ppd MOLA elevation map. Albedo 3 0.15–0.35 20-ppd TES annual maps Thermal Inertia 21 5–5000 Interpolated from temperature. surfaces with lower values of thermal inertia (Fig. 3a). Note that the temperature curves cross one another near dawn and dusk; the derived thermal inertia is non-unique at those times. Figure 3b shows that for ranges typically found on Mars, albedo can affect the surface temperature by ∼20 K, with higher albedo reflecting more solar irradiation and leading to lower temperatures. Atmospheric pressure changes on Mars can affect surface temperature by ∼10 K, with a lower pressure leading to lower temperatures (Fig. 3c).
Martian dust opacities can affect the temperature by ∼10 K, an affect similar in form to that of thermal inertia, where lower dust opacity leads to a larger diurnal temperature range (Fig. 3d).
A north-south slope has a similar effect to changing the latitude in the same direction; e.g. a slope tilted towards the equator will show a similar temperature response to one that is closer to the equator (Fig. 3e).
An eastward facing slope warms up sooner in the morning and heats up more than a westward facing slope
(Fig. 3f).
Mellon et al. (2000) estimated uncertainties of 6% for thermal inertia derived from TES bolometer mea- surements using the earlier model, noting that this value is an upper limit. The major sources of uncertainty come from observations, particularly those of albedo and dust opacity, and how past measurements of these values relate to conditions at the time of subsequent temperature observations.
2.2 Surface Heterogeneity
Each value of thermal inertia derived will be accurate so long as the observation on which it is based sampled a horizontal thermally homogeneous region. In this case, the derived thermal inertia should be identical regardless of whether the observation was taken during the day or night, and regardless of the season during which the observation was taken. However, if the observation footprint covered a region composed of a horizontal mixture of different thermal inertia values, or if it covered a layered region with the upper layer thinner than a seasonal skin depth, then different surface components may have different temperatures at any
8 Figure 3: Model diurnal surface temperatures for (a-d) node values from the lookup table and (e, f) a few ◦ ◦ illustrative sloped surfaces. All models are computed for 0 Ls at 4.6 S, the latitude of the Curiosity landing site. In each panel the dashed line represents an identical model with I = 223 tiu, A = 0.25, P = 5 mbar, τD = 0.0, and a level surface. Thermal inertia nodes are 5, 7, 10, 14, 20, 28, 40, 56, 79, 112, 158, 223, 316, 446, 630, 889, 1256, 1774, 2506, 3540, and 5000 tiu. Albedo nodes are 0.15, 0.25, and 0.35. Surface atmospheric pressure nodes are 0.8, 2, 5, 10, and 15 mbar. Dust opacity nodes are 0.0, 0.4, and 0.8. Sloped models are calculated for (e) an angle of 20◦ with a south-facing azimuth and (f) angles of 0, 10◦ with east and west facing azimuths. Adapted from Putzig & Mellon (2007).
9 given time. The observed brightness temperature will be a mixture of the temperatures of the components within the sample area. Due to the non-linear dependence between temperature and thermal inertia, the derived thermal inertia will not be a linear mixture of component thermal inertias. Evidence from imaging at high resolution indicates that the surface of Mars is heterogeneous on the scales of TES observations and
THEMIS images, and therefore it is generally more appropriate to refer to the derived thermal inertia as the apparent thermal inertia, which is expected to be time and season dependent. Similar effects are expected for surfaces with varying slopes.
The derivation of apparent thermal inertia, rather than the true averaged value for the surface, may appear to represent a drawback in the derivation technique. However, the apparent thermal inertia’s seasonal and diurnal dependence provide a tool which can be used to analyze the sub-pixel and sub-surface structure of the near-surface. Putzig & Mellon (2007) developed a technique that allows seasonal and diurnal variations in apparent thermal inertia to be compared to simple two-component situations. They defined four idealized surface materials (dust, sand, duricrust and rock) with thermal inertia on nodes of the lookup table (to minimize interpolation errors). Typical basaltic values were assigned for density and heat capacity, adjusted to account for a sulfate cement in the case of the duricrust. Albedo values were taken from modal values for Mars (Putzig et al. 2005). The parametrization of the materials is described in Table 3. Seven layered models (dust over sand, crust, and rock; sand over rock; rock over sand; and crust over dust and sand), at a range of upper layer thicknesses, and six horizontal mixes (all possible combinations) at a range of component ratios were produced. The apparent thermal inertia for each of these models was compared with that derived from TES measurements over all available seasons and times of day to infer likely structures for a number of locations on Mars (Putzig & Mellon 2007). I use the same techniques to analyze locations within Gale Crater in Section 4.
Table 3: Idealized Surface Materials
Type Inertia Albedo Density Heat cap. Skin Depth (m) (tiu) (kg m−3) (J kg−1 K−1) Diurnal Seasonal Dust 56 0.26 1375 837 0.008 0.212 Sand 223 0.16 1650 837 0.027 0.702 Duricrust 889 0.23 1875 854 0.093 2.413 Rock 2506 0.16 2500 837 0.201 5.206
10 2.3 Instruments
Observations by the following instruments have been used to provide data for the thermal model during this study; and a description of key instrument parameters are included below.
2.3.1 TES
The Thermal Emission Spectrometer onboard the Mars Global Surveyor was designed to determine the surface mineralogy, volatile abundance and history, and atmospheric dynamics of Mars (Christensen et al.
2001). It made thermal infrared spectral observations between 5.8 and 50 µm, bolometric thermal radiance observations between 5.1 and 150 µm, and visible/near-infrared solar reflectance measurements. It had a spatial resolution of 3 km at the surface of Mars. MGS arrived at Mars on the 11th of September 1997, TES began taking measurements during the aerobraking phase of its mission, and it began the formal mapping phase on the 1st of April 1999. MGS was placed into a fixed-local-time orbit of Mars with equator crossing times at ∼2am and ∼2pm local time. Contact was lost with the spacecraft in November 2006, though the spectrometer instrument on TES failed previously in April 2004. Dust opacity extractions (Smith et al.
2001b; Smith 2004) used in the thermal inertia derivation are accessed from a database of TES data. Albedo is read from global maps produced from TES observations by Putzig & Mellon (2007) and Putzig et al.
(2013). I also used maps of seasonal and annual mean thermal inertia derived from TES measurements which were produced by Putzig & Mellon (2007).
2.3.2 THEMIS
The Thermal Emission Imaging System onboard the Mars Odyssey spacecraft was designed to allow global mapping of unique compositional units at resolutions that permit mineral and rock distributions to be related to geological features (Christensen et al. 2004). It takes visible-spectrum images at 18 m per pixel resolution with up to 5 different color filters, and it takes infrared images at 100 m per pixel resolution with up to 9 different wavelength filters between 6.2 and 15.3 µm. I used radiance data from THEMIS IR band 9, at
12.57 µm, to estimate the brightness temperature of the surface as a first step in calculating thermal inertia.
Figure 4 shows the times of day and the seasons when THEMIS images sampled Gale Crater. Before mid-2009, the Mars Odyssey spacecraft passed over Gale Crater at around 5am and 5pm, with a seasonal oscillation about these times of around 1 hour. After mid-2009, the spacecraft’s orbit was altered and it began passing overhead about 1.5 hours earlier, but with the same oscillation. From Figure 3 it is clear that the majority of THEMIS daytime observations occur near dusk when thermal inertia is non-unique, and for
11 this reason daytime THEMIS images are of little value for the derivation of thermal inertia. Most of the
◦ ◦ THEMIS images that sample Gale Crater were taken when Mars was between 330 and 50 Ls, around the southern autumn equinox, with very little coverage around the spring equinox.
Figure 4: Top: The times of day for THEMIS images which sample Gale Crater. Times for daytime and nighttime images are hours after local noon and midnight, respectively. Before mid-2009 the Mars Odyssey spacecraft passed over Gale Crater at around 5am and 5pm, with a seasonal oscillation about these times of about an hour. After mid-2009 the spacecraft’s orbit was altered and it began passing overhead around 1.5 hours earlier, but with the same oscillation. Bottom: Count of THEMIS images that sampled Gale Crater, binned by season. More observations of Gale Crater have occurred close to the the southern autumn equinox. Nighttime bars are plotted behind partially transparent daytime bars; e.g. during the first 10◦ of the Martian year there have been 3 nighttime and 4 daytime observations.
2.3.3 REMS
The Rover Environmental Monitoring Station (REMS) onboard the Mars Science Laboratory Curiosity rover is designed to investigate the environmental factors that relate to the habitability of the Martian surface
(Gomez-Elvira et al. 2012). It measures air and ground temperature, pressure, relative humidity, wind speed and ultraviolet radiation. The REMS Ground Temperature Sensor (GTS) is located on Curiosity’s mast, and it uses infrared radiometry to measure the temperature in a roughly 100-m2 field of view. The large field of view was chosen to avoid small-scale temperature effects caused by, for example, shadowing by individual boulders. The field of view was designed to minimize the direct heating of the area by the rover
12 itself, though significant thermal contamination still occurs (Gomez-Elvira et al. 2012).
2.3.4 MOLA
In order to scale the Tillman et al. (1993) fit for atmospheric pressure at the Viking Lander 1 site to the appropriate value of locations elsewhere on the planet, it is necessary to know the surface elevation. The
Mars Orbiter Laser Altimeter (MOLA) onboard MGS provided measurements of the elevation of the surface to a vertical accuracy of 1 m (Smith et al. 2001a). The surface spot size for the laser beam from the mapping orbit was 168 m, along-track shot spacing was 300 m, and cross track shot spacing was 4 km at the equator.
I used MOLA elevations binned at 20-ppd in a map created by Putzig & Mellon (2007) for the same purpose.
Maps of slope also created from MOLA data by Putzig & Mellon (2007) are used to adjust the latitude and time of day corresponding to each observation before it is processed using the thermal inertia derivation routine.
2.3.5 Imaging Instruments
For analysis of individual sites within Gale Crater, I used imagery provided by the High Resolution Imaging
Science Experiment (HiRISE) onboard the Mars Reconnaissance Orbiter (MRO), which provides visible- spectrum images at sub-meter resolution (McEwen et al. 2007). Digital terrain models (DTM) created from stereo image pairs provided elevation models with 25-cm vertical precision and horizontal resolution half as detailed as the original images (McEwen et al. 2007). HiRISE monochrome images are taken using red light, and span 6 km with 30-cm resolution at the surface of Mars. The central part of each red image is also sampled by blue-green and near-infrared sensors to provide false-color images with a width of 1.2km at the same resolution. HiRISE images do not cover more than a few percent of the planet’s surface, and DTMs cover even less, but where available they have proven to be invaluable.
The MRO Context Camera (CTX) has a lower resolution and a larger field of view than HiRISE, with which it typically provides complementary observations. It has a resolution of ∼6 m per pixel at the surface
(Malin et al. 2007) and is used to view areas larger than those imaged by HiRISE.
The High Resolution Stereo Camera (HRSC) onboard the Mars Express orbiter provides images and
DTMs at 10–20 m per pixel at the surface (Albertz et al. 2004; Neukum & Jaumann 2004; Jaumann et al.
2007). For this work it was primarily used in areas where HiRISE DTMs were not available.
13 3 The MARSTHERM Website
The MARSTHERM website is located at https://marstherm.boulder.swri.edu. A summary of its capabilities is included on the home page, and public access to the rest of the site will begin on December 12th 2013. The website is written in HTML, PHP and JavaScript, and requested jobs are written to a MySQL database.
The database is monitored by a backend python script to control the processing of individual jobs. The website provides access for registered users to a number of thermophysical analysis tools, described below.
Numerical Thermal Model The flagship component of the website is an interface providing access to the MARSTHERM numerical model described in Section 2. All of the user-defined parameters listed in
Table 1 can be specified in an HTML form. Figure 5a provides a view of the MARSTHERM website that shows the options available for the thermal model. Output can be downloaded as ASCII data files, and in graphical form. Examples of the diurnal and seasonal temperature graphs that are generated automatically by the MARSTHERM website are shown in Figures 5b and 5c.
Global maps Global 20-ppd maps of MOLA elevation, TES-derived albedo, and annual-median and ∆10◦ seasonal TES-derived thermal inertia (Putzig & Mellon 2007) are available to view or download from the website.
Project Area Creation Users can designate an area of interest by creating a project, as described in
Section 3.1. Access to the tools for deriving thermal inertia from THEMIS images and for heterogeneity modeling is provided within the bounds of each project.
THEMIS Processing The MARSTHERM website provides an interface that allows users to specify
THEMIS infrared (IR) images, which are automatically downloaded and processed to derive thermal in- ertia for each pixel, providing output in the form of Hierarchical Data Format 5 (HDF5) data files and georeferenced tiff files. This process is described in Section 3.2.
Documentation Comprehensive HTML documentation is provided to help users understand how to use the tools provided. These documents describe the tools available on the website, explain how to use them, and list the options available for customizing each model. Some example scripts demonstrating how the output HDF5 files can be read using python, MATLAB and IDL are also provided.
14 (a)
(b) (c)
Figure 5: (a) View of the MARSTHERM website showing the thermal-model interface. (b) Diurnal and (c) seasonal temperature variations for assumptions suitable for the floor of Gale Crater. Times are local solar times. These graphs were automatically generated by the thermal model attached to the MARSTHERM website.
15 3.1 Project Designation
In order to process THEMIS IR images, MARSTHERM users are first required to designate a project area.
Each THEMIS image is then trimmed to the bounds of that project to ensure that pixels outside of the
researcher’s region of interest are not unnecessarily processed. Designating a project instructs the system
to generate a number of ancillary data files and plots, some of which are utilized to process the THEMIS
image, and some of which are generated as stand-alone products.
3.1.1 Project Maps
Global maps at 20-ppd of TES albedo, MOLA elevation and TES-derived thermal inertia (Putzig & Mellon
2007) are trimmed to the project bounds. Interpolation from the elevation and albedo maps to THEMIS
pixel locations is carried out during THEMIS image processing.
Three annual albedo maps are generated for each project, since global dust storms changed the albedo of
large regions twice during the period when THEMIS and TES were both operational. These albedo maps are
extracted from those used by Putzig & Mellon (2007) and Putzig et al. (2013) to derive apparent thermal
inertia from TES observations.
A final step in THEMIS image processing is to calculate comparisons between the median thermal
inertia derived from the THEMIS image and from TES measurements for the same region. This is done by
interpolating from the TES-derived maps to each THEMIS pixel location. Comparisons are made between
◦ the THEMIS results and both the (day or night) annual-median maps and ∆40 Ls seasonal maps. The
◦ ◦ ∆40 Ls seasonal maps are generated because data within the ∆10 Ls maps is typically too sparse to allow sensible interpolation to THEMIS pixels.
◦ ◦ Project maps are made available on the website, including all seasonal (both ∆10 Ls and ∆40 Ls) thermal inertia maps.
3.1.2 Dust Opacity
Infrared atmospheric dust opacity was derived from TES daytime observations and stored in the TES
database (Smith et al. 2001b; Smith 2004). Upon creation of a new project these data are used by the
MARSTHERM website to generate a history of dust opacity within the project area. The dust-opacity
values are saved in an array together with the solar longitude and the date and time of the observation. If
the project area is so small that fewer than 2000 suitable observations occurred within the project bounds
during the TES mission, then the project bounds for this purpose alone are expanded automatically. There
16 is no attempt to ensure that the dust-opacity data are evenly distributed by season or time of day, though
ensuring 2000 observations typically results in a reasonable spread throughout all seasons. To approximate
visible dust opacities needed as input by the MARSTHERM model, the TES infrared dust opacities are
multiplied by 2 (Mellon et al. 2000).
3.1.3 Two Component Modeling
The techniques for modeling two-component apparent thermal inertia described in Section 2.2 are used
to produce two-component model curves for the central latitude of the project area, and these curves are
◦ displayed together with median TES-derived thermal inertia values from the ∆40 Ls maps. The comparison will allow users to observe similarities between the models and the TES-observed behavior. The model curves
are also available to download as an ASCII data file.
3.2 THEMIS Image Processing
In the following discussion, the Courier Font is used to denote individual programs. Unless otherwise
stated, all programs are routines available within United States Geological Survey (USGS) Integrated Soft-
ware for Imagers and Spectrometers (ISIS3).
3.2.1 Download and Pre-Processing
The MARSTHERM website automatically downloads georeferenced THEMIS infrared images from the repos-
itory at Arizona State University (http://static.mars.asu.edu). Each image is initially processed to ISIS3
cube format using THMPROC, which projects the camera image to a uniform map, using Mars Odyssey
ephemeris data. The image is then trimmed to the project bounds using MAPTRIM and CROPSPECIAL. The
header of the file is read at this stage to find its spatial extents, solar longitude, exposure date and time, and
the bands available within the image. THEMIS has 10 IR bands, though frequently only 3 bands are included
in the downloaded images. Band 9 centered on 12.57 µm is used for brightness temperatures. Thermal drift correction also requires band 10, centered on 14.88 µm, since the Martian atmosphere is opaque at this wavelength (Christensen et al. 2004). Bands 9 and 10 are always included in the downloaded images. Drift correction is conducted by THMDRIFTCOR. After this step, only the drift-corrected band-9 image remains.
PHOCUBE is used to supplement each pixel in the image with meta data. This process expands the file to
5 bands, which include the drift-corrected radiance value from band 9, the latitude, the longitude, the solar
incidence angle, and the sub-solar ground angle (that is, the bearing from North to the point on the planet
17 which has the Sun at its zenith). PHOCUBE is the final segment of the processing within the ISIS3 software
suite.
The atmospheric dust opacity is taken from the project dust-history file. If the THEMIS image was
taken before the failure of the TES spectrometer in April 2004, then a weighted mean of the 5 dust opacity
measurements that occurred nearest to the THEMIS exposure time is used (weighted by time proximity). If
the THEMIS image was taken after the failure of the TES spectrometer, then the median value for all dust
◦ opacity measurements taken within 0.5 Ls of the exposure solar longitude are used, ignoring measurements taken during the global dust storm of Mars year 25. To emphasize:
• For early THEMIS images, actual measurements of TES opacity at or near to that time are used.
• For later THEMIS images, it is assumed that opacity is seasonally dependent, and a median TES value
for that season is used.
Figure 6 shows an example dust opacity history for a project centered on Gale Crater. There is considerable
scatter around a seasonal trend. The option to override the infrared dust opacity selected using this method
with a custom value is available on the web interface.
Elevation and albedo are interpolated from the MOLA and TES project maps to each THEMIS pixel
position. The atmospheric pressure at the Viking Lander 1 elevation for the image Ls is found using the fit from Tillman et al. (1993), and is then scaled to the elevation of each pixel. Dust opacity for each pixel
is then scaled to the appropriate pressure. The TES albedo map used depends on the date of exposure for
the THEMIS image. If the image was taken before the end of the MGS TES mission, then the albedo map
for the corresponding Mars year is used. Otherwise the newest albedo map is used. The effects of any dust
storms since the failure of the TES instrument are not considered.
THEMIS band-9 radiance is converted to brightness temperature (Tb) by interpolation from a lookup table generated using the make temp rad routine from the Da Vinci software suite, as described by Christensen
(2003). Latitude is known for each pixel, and Mars day is derived from Ls. Mars hour must be calculated from the pixel latitude and longitude (φpix, θpix), the solar incidence angle (λ), and the sub-solar ground azimuth (Ψ). First the sub-solar latitude and longitude (φsol, θsol) are calculated:
sin (φsol) = sin (φpix) cos (λ) + cos (φpix) sin (λ) cos (Ψ) (6)