By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 A number of physical properties experiments will be conducted during the NASA 2003 Mars Exploration Rover Mission as the two vehicles explore Meridiani Planum and the floor of Gusev Crater. The investigations will include quantifying dust accumulation and dispersal dynamics by periodically monitoring the rover decks with the Athena Pancam and Mini-TES instruments. Properties of soil-like materials will be inferred from analyses of wheel track patterns, depths, and wheel slippage dynamics during traverses. The rovers will be modeled as dynamic mechanical systems to extract along-track terrain topography and physical properties from times series of rover tilt vectors, wheel encoder counts, azimuths, motor currents, and rocker and bogie angles. Trenches will be excavated using rover wheels to characterize mechanical properties of soil-like materials with depth and to expose subsurface materials for remote and in situ observations using the Athena Payload. The Rock Abrasion Tools will be used to expose rock subsurfaces for detailed analyses. Motor currents and penetration magnitudes will be compared to a database of rocks excavated by an engineering model of the Rock Abrasion Tool to understand Martian rock mechanical properties. Image-based localization analyses will be pursued to better understand rover traverse directions and magnitudes and thus rover locations as a function of time. The physical properties and localization investigations, when combined with analyses of the full ensemble of Athena observations, will greatly improve our understanding of Martian surface properties and provide significant technology lessons for future landed missions.
 The intent of this paper is to describe a set of physical properties and localization investigations that will utilize the NASA 2003 Mars Exploration Rovers (MER) [Crisp et al., 2003] and associated Athena Instrument Payload [Squyres et al., 2003] during the period in which the vehicles are exploring Meridiani Planum and Gusev Crater. The work is focused on increasing our understanding of terrain properties and providing knowledge and experience that will be used to enhance operations during future landed missions (Tables 1 and 2). The paper proceeds by first briefly summarizing existing knowledge about the nature of Martian surface materials, followed by an overview of the rover capabilities and how the experiments will be accomplished within the context of overall mission objectives. An overview of image-based localization studies for science and mission operations is then presented. These sections are followed by a discussion of monitoring aeolian dust deposition and erosion by imaging the rover decks, calibration targets, and magnets. Remaining sections focus on how wheel tracks, wheel slippage, and trenching will be used for geological and physical properties studies, followed by an overview of how the rover will be modeled as a dynamical system to extract along-track estimates of terrain properties. The paper ends with a section on how the overall suite of Athena Payload observations will be used together with the rover-based experiments to address unanswered questions about Martian surface properties (Tables 1 and 2), in addition to providing lessons to be used to enhance operations for future landed missions.
Table 1. Mars Exploration Rover Physical Properties and Localization Science Objectives and Approaches: What are the Characteristics, Relative Ages, and Origins of Surface Materials on Mars, Including Bedrock, Blocks, and Soil-Like Deposits?
*Determine if bedrock outcrops, strewn block fields, duricrust (blocky soil-like materials), and wind-blown deposits at the MER sites that are similar to those found at the Viking and Pathfinder sites or if there are new surfaces and materials not yet encountered.
*Use rovers to acquire remote sensing and in situ observations from various positions, with image-based localization employed to accurately define traverses.
*Generate maps for bedrock and surficial units and compare to Viking and Pathfinder-based maps.
*Determine areal distribution and stratigraphic relationships among mappable rock and soil-like units.
*Analyze maps and remote sensing data to establish stratigraphy.
*Use rover wheel-trenching procedures to excavate subsurface soil-like materials and determine microstratigraphy and physical properties for soil-like units.
*Determine properties for each type of material (e.g., grain size frequency and angularity, density, mineralogy, bearing strength, cohesion, angle of internal friction, slope distributions).
*Use ensemble of Athena observations to advantage, including photometry from Pancam, emissivity and thermal inertia estimates from Mini-TES to infer properties for rocks and soil-like units.
*Compare estimates to those for the Viking and Pathfinder sites.
*Image and analyze wheel track patterns from traverses to infer grain sizes, compaction, and bearing strength.
*Conduct trenching experiments to infer cohesion and angle of internal friction with depth in soil-like materials.
*Utilize rover traverse telemetry and dynamic models to infer terrain slope distributions and soil properties.
*Determine if characteristics and distribution of materials can be correlated with the orbital remote sensing observations, including estimates of surface roughness, reflectivity at various wavelengths, thermal inertia, block distributions, composition, and mineralogy.
*Use all rover-based observations to synthesize orbital views for comparison to actual orbital data.
*Conduct selected experiments where rover-based and orbital observations are acquired simultaneously (e.g., Mars Express OMEGA, Athena Pancam and Mini-TES.
*Determine bedrock and surficial geologic histories by combining the physical properties experiments with the overall ensemble of Athena Payload and orbital observations.
*Synthesis of available data, including results from specific physical properties experiments.
*In particular, determine history of interaction of crustal materials and water, using evidence from bedrock and soil deposits.
Table 2. Mars Exploration Rover Physical Properties and Localization Technology Objectives and Approaches: How Will the Physical Properties Experiments Provide Feed-Forward Surface Property Information and Technologies for Future Landed Missions?
*Quantify unit properties, e.g., grain size, density, cohesion, angle of internal friction, bearing strength.
*Analyze traverse telemetry data, specific trenching experiments, and relevant Athena observations to infer properties of rock and soil-like units.
*Consider implications for future landed missions, e.g., nature of dust deposits and extent to which the material will degrade spacecraft power (for solar-powered systems) and mechanical systems (e.g., motors and bearings) and thus shorten mission lifetimes.
*Monitor dust accumulation and dispersal on deck and other surfaces using Pancam and Mini-TES and evaluation of solar panel power output.
*Consider properties of dust and extent to which the material impacts operations overall.
*Extrapolate findings to other sites to infer terrain types, suitability for landing and use of mobility systems (both lateral, i.e., rovers, and vertical, e.g., drills)
*Evaluate hazards of relevance to human surface missions, e.g., ubiquitous and potentially hazardous dust deposits.
*Utilize ensemble of experiments and observations to consider impacts on human missions, including ability to excavate and shape soil-like deposits, extract relevant materials, what might be hazardous (e.g., heavy metals in dust), and atmospheric disturbances such as dust devils.
*Consider what materials can be used to advantage during a human mission, e.g., for shielding from radiation or processing to extract water or other needed subcomponents of surface units.
*Quantify how surface activities can be used to plan ahead for more autonomous surface systems for future landed missions, e.g., quantitative modeling of rover dynamics and localization, with on-board feedback loops for long-distance traverses with successful goal acquisition.
*Conduct focused localization experiments, using overlapping images to track features, images of wheel tracks, and wheel currents and turns to fuse quantitatively into best estimates of rover navigation. (Image based localization by tracking features is not part of on-board capabilities so work would be with archival data.)
*Evaluate capabilities and limitations of as a mobility system for traversing landing site units.
*Evaluate capabilities and limitations of the Instrument Deployment Device for deploying assets onto targets.
2. Global Setting and Current Knowledge of Martian Physical Properties
2.1. Global Views
 Global scale albedo and color contrasts recorded during decades of telescopic monitoring provided the earliest clues that Martian surface materials are heterogeneous and complex, with long-term bright and dark regions that cover hundreds to thousands of kilometers in lateral extent and dust storms that range from local to global in scale [e.g., Slipher, 1962]. Telescopic and orbital spectroscopic measurements suggest that the bright regions are covered by fine-grained (i.e., dust), ferric-oxide-rich cryptocrystalline glasses and associated devitrification products generated by alteration of basaltic minerals during impact events, hydrothermal activity, and perhaps during low-temperature aqueous alteration [e.g., Morris et al., 2000, 2001a, 2001b]. Spectral properties of dark regions are interpreted to be basaltic rocks, perhaps with basaltic andesites in the northern lowlands [Bell, 1999; Christensen et al., 2001a; Bandfield, 2002]. Thermal inertia estimates are consistent with these interpretations in that bright areas are inferred to be dust-covered whereas dark areas are interpreted to be covered with sand-sized or coarser deposits [Mellon et al., 2001] (Figures 1 and 2). Rock abundances have been estimated by modeling spectral emission as a combination of dark rocks with a fixed inertia and bright soil-like materials [Christensen, 1986; Golombek et al., 2003]. The models solve for the areal fraction of rocks and the inertia of the soil-like materials. Results follow the distribution of bright and dark areas closely, with a higher areal fraction of rocks exposed on low-albedo terrains. On the several-kilometer spatial scale of the Mars Global Surveyor Thermal Emission Spectrometer (TES) observations, extensive exposures of bedrock are very rare [Mellon et al., 2001]. On the other hand, high spatial resolution thermal mapping (approximately 100 m/pixel) by the Odyssey Thermal Emission Multispectral Imaging System (THEMIS) instrument shows that some dark areas are dominated by bedrock exposures (P. R. Christensen, personal communication, 2003).
 Detailed analyses of global-scale albedo and thermal inertia data demonstrate that on a global basis there is a third surface type in addition to bright and dark areas [Arvidson et al., 1989a; Mustard and Cooper, 1999; Mellon et al., 2001]. This surface type is red and has albedo and thermal inertia values intermediate of those for bright and dark areas (Figures 1 and 2). Further, examination of imaging data shows that bright aeolian streaks are often superimposed on this surface, along with dark dune deposits [Arvidson et al., 1989a]. One interpretation is that these intermediate albedo, red surfaces expose indurated soil-like materials that are represented at the two Viking Lander and the Pathfinder landing sites by duricrust and cloddy to blocky soil-like materials [Moore et al., 1987, 1999].
2.2. Viking and Pathfinder Landing Sites
 Viking Lander 1 touched down on the surface of Mars on July 20, 1976 in Chryse Planitia [Binder et al., 1977] whereas Viking Lander 2 touched down on Utopia Planitia on September 3, 1976 [Mutch et al., 1977]. We focus discussion on the Viking Lander 1 results for sake of brevity and because similar surface materials were sampled at both Viking sites. The Viking Lander 1 site is interpreted from orbital observations to be a cratered volcanic plain with a suite of curvilinear ridges striking approximately north-south and of probable tectonic or volcanic origin [Binder et al., 1977] (Figure 3). From the surface some bedrock can be observed, along with a number of discrete rocks, most of which were probably emplaced as crater ejecta (Figure 4). Indurated soil-like materials were identified in many locations, including cloddy to blocky materials and polygonal ground beneath the retrorockets where fine-grained drift material was eroded by rocket exhaust (Figure 5). On a site-wide basis, aeolian drift material is superposed on and mixed with blocky to cloddy materials (Figures 5, 6, and 7). Several dark deposits were also identified that are interpreted to consist of windblown sands (Figure 4). During the three Martian years of observations from Viking Lander 1 the surface was observed to brighten after each dust storm, only to slowly return to pre-storm conditions. Several aeolian erosional events were also witnessed during which soil-like clods were eroded and removed from the scene [Arvidson et al., 1983].
 Surface materials exposed at the Viking Lander 1 and Pathfinder sites are generally consistent with albedo and thermal inertia estimates derived from orbital observations. The Viking Lander 1 site has fewer dark rocks and more drift material, consistent with the higher albedo of this site as opposed to the Pathfinder site (Figure 1). Likewise the thermal inertia of the Pathfinder site is higher than found for the Viking Lander 1 site, consistent with the greater areal density of rocks at the former site (6% areal coverage for Viking Lander 1 as opposed to 20% for Pathfinder for rocks larger than 3 cm in diameter [Golombek et al., 2003]). Both sites are anomalous relative to the global trend in which there are three clusters in albedo-thermal inertia space: bright, low-inertia dusty areas; dark, high thermal inertia surfaces (presumably with an abundance of sand-sized materials); and the intermediate surface thought to result from exposures of duricrust (Figures 1 and 2). Both the Viking Lander 1 and Pathfinder sites have albedos that are high relative to the thermal inertias. Rocky surfaces with a thin cover of drift material would explain the observations, a hypothesis consistent with available lander and rover-based data.
2.3. Viking and Pathfinder Physical Properties Experiments
 Mechanical properties of Martian surface materials have been measured by all three successful Mars soft-landing spacecraft and provide data that are complementary to the orbital observations and geologic setting information presented in the last section [Moore et al., 1977, 1978, 1982, 1987, 1999; Arvidson et al., 1989b; Rover Team, 1997b]. The two Viking Landers utilized extendable surface sampler arms to dig into Martian soil-like material, to reveal and probe surface crusts, and to push against rocks (moving some of them) (Figures 4 and 6). Transformation of arm actuator currents and image measurements into standard soil mechanical parameters such as cohesion, c, angle of internal friction, ϕ, and bulk density values required application of a plowing analogy for some activities of the soil scoop [McKyes and Ali, 1977]. Trench walls, slopes of artificial piles of material, and footpad imprints were also analyzed to determine mechanical properties of surface materials [Moore et al., 1977, 1982, 1987; Arvidson et al., 1983].
 Drift, blocky, and crusty-to-cloddy soil-like materials were identified at the Viking sites (Figure 6). Drifts had the lowest ϕ values (∼18°) and were interpreted to be very fine-grained deposits according to results of the Viking Gas Exchange Experiment [Oyama and Berdahl, 1977; Ballou et al., 1978]. For blocky material, ϕ typically was ∼30°. Crusty-to-cloddy material, thought to be disrupted duricrust deposits, had ϕ values of about 35°, while ϕ was ∼31° for mixed fines and crusty materials. Cohesions ranged from 0.7–3.0 kPa for drift, 1.5–16 kPa for blocky material, 0.5–5.2 kPa for crusty-to-cloddy material, and 0.2–2.3 kPa for mixed drift material and crusts. Estimates of bulk densities ranged from about 1200 kg/m3 for drifts to 1600 kg/m3 for blocky material (results summarized by Moore et al.  and Arvidson et al. [1989b]). Rocks were too strong to be chipped or scratched by the surface sampler [Moore et al., 1977, 1978]. Moore et al.  concluded that cohesive strengths characteristic of surface materials at the landing sites explained the stability of these materials against mobilization by winds, in contrast to the vulnerability of artificial piles of material and other areas of disturbed soil-like material.
 The Pathfinder lander did not have a surface sampler arm and used airbags instead of foot pads to cushion its landing, so soil mechanical properties were derived mainly from analysis of rover wheel tracks and wheel-trenching sites (Figure 8). Analysis of rover suspension telemetry, frictional wheel torques indicated by motor currents, and images of disturbed soil-like material yielded values of ϕ and c, and clues about particle size-frequency and bulk density to depths of a few centimeters beneath the surface [Rover Team, 1997b; Moore et al., 1999]. Interpretation of data was aided by ground-based experiments measuring the behavior of rover hardware in two test sands and a lunar simulant, although ϕ and c values measured by conventional laboratory methods were not obtained [Moore et al., 1999].
 The main technique for determining ϕ and c values with the rover involved holding five wheels fixed while rotating the sixth, and monitoring frictional resistance of the rotating wheel as it dug into the soil-like material [Rover Team, 1997a, 1997b; Moore et al., 1999] (Figure 8). At the beginning of each wheel-trenching, contact surface area between the wheel and the ground was small, maximizing normal stress; contact surface area increased and normal stress decreased as the wheel dug deeper. Matched pairs of shear and normal stress values recorded during each wheel-trenching were plotted and fit by straight lines to determine ϕ and c. Bulk densities and porosities could not be measured directly, but were estimated on the basis of approximate correlation with ϕ noted in studies of lunar regolith [Mitchell et al., 1972].
 Several material types were identified at the Pathfinder site on the basis of mechanical properties, color, and morphology, and they closely match those observed at the Viking sites. Drift material (ϕ = 15.1° to 27.9°, c = 0.18 to 0.53 kPa) was distributed around the landing site in thin surface deposits, usually covering cloddy and rocky materials. High-fidelity castings of wheel treads in drift material indicated a substantial component of silt- and/or smaller-sized particles [Rover Team, 1997b; Moore et al., 1999]. Cloddy deposits (ϕ = 31.4° to 42.2°, c = 0.18 to 0.53 kPa) consisted of poorly sorted materials including granule-sized clods and/or rock grains, and exhibited crust-like behavior in some locations [Rover Team, 1997b; Moore et al., 1999] (Figure 8). Other, less common materials were also encountered. “Scooby Doo,” an apparently indurated deposit, was too coherent for strength parameters to be determined using the limited normal stresses available from the rover wheel. “Mermaid,” a low dune-like feature, was traversed by the rover with somewhat ambiguous results regarding particle size and internal structure, but mechanically these materials are similar to cloddy material. Negative cohesion values derived from linear fits to some data indicate the challenges involved using the rover's wheel/suspension system to measure and resolve low cohesion values of Martian soil. Nevertheless, in some places changes in wheel behavior and digging efficiency with depth indicated excavation into multiple mechanically distinct layers (e.g., through drift material into cloddy material) and these relationships were confirmed by evidence from images [Moore et al., 1999].
3. Rover Description
 The two MER rovers are identical mobility systems and each carries an Athena Instrument Payload and sets of engineering cameras [Crisp et al., 2003; Squyres et al., 2003] (Figure 9 and Table 3). Each rover has a mass of approximately 180 kg. At the rover wheelbase, each vehicle is approximately 1.4 m long and 1.2 m wide. At the solar panel each rover is 2.25 m wide by 1.5 m long. In deployed configuration (Pancam mast assembly deployed), each rover is just over 1.5 m tall and has a ground clearance of about 0.3 m. The rovers are 6-wheel drive, 4-wheel steered vehicles with a rocker-bogie suspension system similar in design to Sojourner rover. The center of mass is near the pivot point of the rocker bogie system. As a consequence, the vehicles will be able to withstand a tilt of 45 degrees in any direction without over-turning, although fault protection limits prevent the vehicles from exceeding tilts of 30 degrees during traverses. The rover rocker-bogie design allows the traversing of obstacles of approximately a wheel diameter (25 cm) in height. Each wheel has cleats and is independently actuated and geared, providing for climbing in loose soil-like materials and traversing over rocks. The front and rear wheels are independently steered, allowing the vehicles to turn in place and execute arc turns. The rovers have a top speed on flat hard ground of 4 cm/s, but under autonomous control with hazard avoidance, the vehicles will travel more slowly. Each rover carries a Litton LN-200 inertial measurement unit (IMU) that provides 3-axis rate and 3-axis tilt information.
Table 3. Athena Payload and Engineering Camera Definitions
Pancam: Panoramic Camera
Twelve bands (0.4 to 1.0 μm) for stereoscopic imaging with 0.28 mrad IFOV; 16.8 deg by 16.8 deg FOV. Stereobaseline separation of 30 cm. External calibration target on rover deck.
Mini-TES: Thermal Emission Spectrometer
Emission spectra (5 to 29 μm, 10 cm-1 resolution) with 8 or 20 mrad FOV. Internal and external blackbody calibration targets.
IDD-Mounted In Situ Package
APXS: Alpha Particle X-Ray Spectrometer
244Cm alpha particle sources, and x-ray detectors, 3.8 cm FOV.
MB: Mössbauer Spectrometer
57Fe spectrometer in backscatter mode; Co/Rh source and Si-PIN diode detectors; field of view approximately 1.5 cm2.
MI: Microscopic Imager
30 μm/pixel monochromatic imager (1024 × 1024) with 6 mm depth of field.
RAT: Rock Abrasion Tool
Tool capable of preparing 5 mm deep by 4.5 cm wide surface on rocks.
Located front of rover within Pancam FOV. Weak magnet to cull suspended particles from atmosphere.
Located front of rover within Pancam FOV next to Capture Magnet. Strong magnet to cull suspended particles from atmosphere.
Located next to Pancam calibration target. Intended to separate magnetic from non-magnetic particles. To be examined by Pancam.
Four magnets of different strengths in RAT. To be examined by Pancam when IDD points RAT toward cameras.
Navigation Cameras (Navcam)
Mast-mounted panchromatic stereoscopic imaging system with 0.77 mrad IFOV; 45 deg FOV, and 20 cm stereobaseline separation. For planning sequences.
Hazard Avoidance Cameras (Hazcam)
Front and rear-looking panchromatic stereoscopic imaging systems with 2 mrad IFOV; 123 deg FOV, 10 cm stereobaseline separation. For path planning and hazard avoidance during traverses.
 The rovers are expected to drive to a number of locations and then to a number of targets at each location for investigation and measurement by the science payload (Tables 1–3). The rovers are each expected to receive a single command sequence at the beginning of each sol and safely carry out the execution of commands in this sequence. In executing commands that move the vehicles, the rovers will perform traverses over specified distances, precision movements which position the vehicles with respect to targets, and deployments of in situ instruments using the Instrument Deployment Device (IDD). The rovers will automatically estimate their position using wheel odometry and IMU data. The Pancams will be used to provide an absolute heading update when needed by imaging the Sun, downlinking the data for analysis, and uplinking a new heading.
 The rover software provides both low-level navigation commands to move or turn the rovers as well as high-level autonomous functions that will look for and avoid hazards automatically during traverses. The “go_to_waypoint” command contains a coordinate parameter giving distance and heading of the intended destination of the vehicles, a tolerance of accepted distance near the destination, a distance that quantifies missing the targets, and a timeout parameter for completion of the drives. The rovers will perform hazard detection during traverses using the Hazcam cameras positioned on the vehicle bodies (Figure 9). While the rovers are stopped, images from these cameras will be captured and processed into local digital elevation maps. These models will be used to determine if the terrain features represent obstacles. A small number of short potential paths in the direction to the destination will be developed within these models and safe paths that avoid obstacles will be chosen on-board. The rovers will move a short distance (about 30 cm) along the paths and the process will then be repeated.
 As the terrain models are acquired, they are organized into “world” models with each rover at the center of an approximately 10 m by 10 m area. Once constructed, the larger models provide more options for safe traverses to destinations and prevent the rovers from encountering obstacles already avoided during prior segments of drives. The rovers proceed to move to the destinations until the tolerances of accepted distances near the destinations are achieved, or (in failure cases) either the rovers have driven distances further than the targets or for time periods that exceed designated values for the traverses.
 Hazcam images collected for use in on-board hazard avoidance can be downloaded for Earth-based analyses, if needed. In addition, the option exists to download engineering files collected during traverses and other activities that includes 8 Hz sampling of wheel encoder counts, azimuthal headings, motor currents, along with rocker and bogie suspension angles, and rover tilt direction and magnitude.
4. Physical Properties Experiment Objectives Within the Context of the Mars Exploration Rover Mission
 All experiments described in this paper will be performed within the overall context of the MER mission and the Athena science investigation [Crisp et al., 2003; Squyres et al., 2003] (Table 3). The Athena investigations have many goals and one of the challenges is to carry out physical properties experiments in ways that help maximize the overall science return of the full investigation (Tables 1 to 3). Operationally, physical properties and technology experiments fall into three categories: (a) experiments that involve no special operations of the rover mobility system or IDD, (b) targets of opportunity, and (c) focused experiments.
 In the first category, rovers and payloads will be used to observe features and phenomena that are the consequence of normal rover operations (Tables 1 and 2). These experiments will typically involve the use of remote sensing instruments. For example, stereo Pancam images of wheel tracks will be used to infer soil-like material properties from track depths. Mini-TES data will be used to model thermal inertia values for wheel tracks and adjacent areas and thus provide information about the degree of compaction caused by the wheels. Pancam color imaging of the rover decks and solar arrays will provide information about the rate of dust buildup from atmospheric fallout and any subsequent erosion. Engineering sensors on the rover mobility systems will provide information concerning the dynamical interactions between the rovers and the terrain as driving occurs, which in turn will provide information about the physical properties of rocks and soil-like materials (Tables 1 and 2). All of these experiments have the desirable characteristic that they can almost be performed “for free”. They do not impact other operations beyond the collection and transmission of the observations, which typically serve multiple scientific purposes.
 A second category of physical properties experiments consists of operations that are performed with the mobility system or the IDD on “targets of opportunity”. Each time that a rover moves, it will proceed toward its intended target or in its intended direction with some amount of error. Because of navigation errors, it is often impossible to predict just where the wheels will be at the end of a drive, or exactly what materials will be found within the work volume of the IDD. As a rover is being directed to a specific target, it may take several sols before that target is reached. However, along the way materials of some interest may be found, by good fortune, to be under a rover wheel, or within the work volume of the IDD. These materials are called targets of opportunity. They may be of somewhat lesser interest than other targets, but they nevertheless provide attractive opportunities because no extra sols must be used to reach them.
 At the end of a rover drive, images will typically be acquired for the scene in front of the rover with the front Hazcams and with the Navcam. These images together document the IDD work volume and the appearance of the materials on which the front wheels rest. The images will normally be downlinked via direct-to-earth communication at the end of the sol, so that they can be used to plan activities for the next sol. One of the first activities might be a Microscopic Imager (MI) deployment onto soil-like materials, i.e., as a target of opportunity measurement. Also, the Mössbauer Spectrometer (MB) contact plate can be pressed into soil-like material and then imaged to study compaction. One of the front wheels can be used to dig a shallow trench, and then the rover can back out of this trench and investigate the excavation and soil-like material pile with remote sensing instruments such as Pancam (to document trench morphology) and Mini-TES (to study the exposed subsurface mineralogy). These operations consume little time, allowing them minimal impact on activities for that sol. Front Hazcam images are required to plan any IDD placement, and there is only one downlink/uplink cycle per sol. For these reasons, IDD placement within a trench must be performed first on the sol following the trenching exercise. Targets of opportunity are generally not appropriate for a full sol activity, so the IDD will not normally be used to investigate a trench that has been excavated in such a target.
 The third category of focused physical properties investigations includes ones performed on carefully selected targets. These are rocks or patches of soil-like material that are determined on the basis of remote sensing data to offer particularly important scientific opportunities (Tables 1 and 2). Because of their importance, it may take multiple sols to get rover wheels onto these targets, or to get the rocks or soil-like materials within the work volume of the IDD. Once this task is accomplished and the IDD or the mobility system is used to interact with the targets, the full suite of Athena payload instruments will normally be used to characterize the material properties of the rocks or soil-like materials of interest.
 For rocks, the primary means for investigation of physical properties will be the Rock Abrasion Tool (RAT). The accompanying paper by Gorevan et al.  describes how RAT engineering data can be used to infer rock physical properties. The full suite of instruments will normally be used on a rock before and/or after use of the RAT, both to document the physical effects of the abrasion, and to study new rock textures and compositions that may be exposed.
 For high-priority soil-like material targets, a rover wheel will be used in a variety of ways for trench excavation, and the full payload can be used to investigate the trench and the pile of soil-like material before and after excavation. After a trench has been excavated, the rover can be positioned such that the trench floor is placed within the work volume of the IDD. If end-of-sol front Hazcam images show that this positioning is successful, on the next sol the APXS, MB, and/or MI can be used to investigate the materials on the floor of the trench or soil-like material pile, adding to the information that would be gained by remote sensing alone. These targeted physical properties investigations will consume considerable time and resources, so they will only be performed on materials of particularly high scientific interest.
5. Image-Based Localization Experiments
 Detailed knowledge of rover location as a function of time will be critical for placing local observations in the context of the overall suite of traverses and measurements and to link these observations to orbital measurements (Tables 1 and 2). UHF Doppler tracking will provide the rover location in an inertial coordinate system with an accuracy of approximately 100 m, possibly 30 m, within several sols after landing [Folkner et al., 1997; Li et al., 2003]. This information will be used to locate the landing sites in orbital images so that mission and path planning can be accomplished within a regional-scale context. Error estimates for the placement of the landing site within an orbital image range as high as several hundred meters. The identification of terrain features seen both from initial Pancam panoramas and orbital images will be used to define the rover location more precisely. It is also critically important to track the rover location relative to the landed location so that the connection can be maintained between what is observed and measured on the surface and the regional-scale context obtained from orbital perspectives.
 The nominal rover localization error during traverses is associated with on-board navigation sensors (via wheel encoder counts) and is about 10% or about 10 m per 100 m of traverse distance. These errors propagate with increasing traverse distance, significantly affecting knowledge of rover locations relative to the landing site and previous positions occupied by the rovers. To compensate for sensor errors an image-based rover localization approach will be employed, featuring a network of surface images and bundle adjustment techniques applied to data received on Earth [Li et al., 2003]. The techniques will also form a basis for on-board localization analyses for future landed missions, such as the Mars Science Laboratory (Table 2).
 The planned experiments include (a) creation of incrementally expanding, high-precision landing site cartographic products (panoramic maps, digital terrain models, orthoimages, and rover traverse maps) based on the bundle-adjusted Pancam, Navcam and Hazcam images (Table 3) that are collected throughout the mission, (b) long range rover localization through incremental and integrated bundle adjustments to achieve an accuracy of approximately a meter when the rover has traveled half a kilometer from the landing site, and (c) linking surface and orbital images using bundle adjusted image networks.
 The objective of bundle adjustment work is to place rovers and terrain features in a coordinate system centered on the landers. Pancam, Navcam, and Hazcam images will be used to build an image network progressively as a rover traverses away from the landing site (Figure 10). Tie points that are image features appearing in the Pancam, Navcam, and Hazcam stereo image pairs and cross stereo image pairs will be selected to refine the camera exterior orientation (EO) parameters using bundle adjustment techniques developed specifically for landed missions [Li et al., 2003]. Because EO parameters are estimated within the entire image network, the images are georeferenced in a higher accuracy across the landing site than done by dead reckoning based on pointing and location information provided in telemetry. This is especially important for images farther away from the landing center. As noted, after bundle adjustment, the refined EO data and Pancam and Navcam images will be employed for generation of map products covering the landing site and rover traverses. The analysis will also utilize location and heading information of the rovers reported in telemetry and derived from wheel turn and other information. The output will be improved knowledge of the path traversed by the rovers, information that will be fed back into mission planning activities.
 The selection of tie points in image margins and overlapping areas, especially in critical forward and back-looking images (i.e., viewing from opposite directions), is a challenging task. We have developed a technique based on a Foerstner Operator [Foerstner, 1986] that can automatically select widely dispersed tie points in overlapping images from terrain features such as rocks. The search for matched features will be constrained to small areas based on epipolar geometry using the initial EO data. The surface rover image networks will be built progressively and the bundle adjustment will be conducted in an incremental manner. First, a 360° panorama of Pancam/Navcam stereo images acquired while the rovers are still on the landers will be tied together to form circular image networks (Figure 10). For each site this early image network will be expanded as the rover collects more data. In order to process the images and to support science and engineering operations in a timely manner only incremental bundle adjustments will be performed on a daily basis, using images returned most recently (e.g., one sol). Every few sols there will be an integrated bundle adjustment for the entire image network. This network will be adjusted globally and is expected to produce the best bundle adjustment result [Li et al., 2003].
 An error propagation model will be developed and applied in which errors in observations are described as a covariance matrix of all estimated unknowns in the adjustment, including tie point position and camera orientation parameters. These errors will be propagated through matrices representing the geometric configuration of the image network. Accordingly, errors of individual parameters and their correlations will be depicted in relevant submatrices of the covariance matrix as variances and covariances. Thus location errors will be given by a 3 × 3 covariance submatrix, from which an error ellipsoid can be derived. The error ellipsoid will reveal error magnitudes and orientations by its three principal axes. Similarly, errors associated with camera positions and orientations can also be extracted from the covariance matrix. This approach will provide a quality index for the subsequent photogrammetric measurements and landing site mapping products.
 As part of a proof-of-concept work on localization we conducted focused experiments using: a. the FIDO rover, a helicopter-borne imaging system, and a simulated rover stereo imaging system at Silver Lake, CA in 1999 and 2000 [Ma et al., 2001; Li et al., 2002], and b. ten Mars Pathfinder IMP stereo image pairs and two Sojourner rover stereo image pairs [Di et al., 2002]. The Silver Lake image network consisted of 76 rover images acquired at 18 stations along a traverse of 850 m. To form the image network, 217 tie points were selected to form the image network. After bundle adjustment, the standard deviation of the rover locations relative to a fixed location was found to be ∼1 m. For the Pathfinder experiment we found that standard deviations in rover camera locations were only about 2% of the distance from the lander to the rover. As noted, for the MER mission, the rover localization accuracy using the image-based localization approach is expected to achieve an accuracy of ∼2 m/km.
 We will also perform an extended bundle adjustment experiment to link orbital images, including Mars Global Surveyor (MGS) MOC, Viking Orbiter Visual Imaging System, Odyssey THEMIS, Mars Express High Resolution Stereo Camera images, and MGS Mars Laser Altimeter (MOLA) altimetric data with the surface-based images in the same coordinate system. The result will be a suite of maps that will allow synergistic analyses of orbital and rover-based observations.
6. Rover Deck Dust Accumulation and Dispersal Experiments
 The most important reason for monitoring airborne dust accumulation on MER rover decks is the accompanying decrease in efficiency of power generation resulting from coating of the solar cells (Figure 9). In fact, dust accumulation could reduce the scope of rover mobility and science operations and could potentially shorten the mission. We expect dust accumulation to be a power issue because, during the Pathfinder mission, the slow decrease in efficiency of power generation was coupled with deposition of aeolian dust on the solar power arrays [Landis and Jenkins, 1997, 2000; Ewell and Burger, 1997]. Accumulation of dust will also change the spectral properties of the Mini-TES and Pancam calibration targets and thus change the method by which these instruments undergo in-flight radiometric calibration. In addition, magnets are mounted on the deck and elsewhere and monitoring these surfaces will provide information on the magnetic phases associated with windblown dust [Madsen et al., 2003]. Finally, monitoring the deck and noting the patterns of dust will potentially provide not only depositional information but also clues as to the extent to which the dust has been remobilized by winds.
 The spectral properties of rover deck hardware (including solar cells, Pancam and Mini-TES calibration targets) will be measured before flight for wavelength regions covered by the Pancam and Mini-TES instruments (Table 3). In addition, the measurements will be repeated using dust covers of varying thickness. Because samples of Martian dust are not available with which to conduct laboratory measurements, experiments using Martian analogs are the only viable approach to estimate the relationship between dust thickness and spectral properties. A suitable dust analogue is the <5 μm size fraction of palagonitic tephra formed under cold, dry conditions. These samples are analogous to Mars with respect to a ferric-rich basaltic composition, spectral properties at visible and near-IR wavelengths, magnetic properties, and size fraction [Morris et al., 2001a]. Several laboratory studies directed at quantifying the relationship between dust thickness and spectral properties for palagonitic dust on rock substrates have been published in the literature [e.g., Johnson and Grundy, 2001; Graff et al., 2001; Morris et al., 2001b] and show that about 300 μm of accumulated dust is infinite thickness with respect to penetration of light to solar array surfaces. Our experiments will focus on measurements up to and including this dust cover thickness.
 During the mission, the Pancams will acquire multispectral images of rover deck surfaces (solar cells, Pancam, and Mini-TES calibration targets) at regular intervals and at comparable sun angles. The Pancams will be in good focus on the decks, permitting good spatial characterization of the dust, with spectral characterization to be defined using the instrument's multispectral capabilities. The Mini-TES instrument will also be used to acquire spectra of solar cells and the Mini-TES calibration targets. Although the Mini-TES is not in good focus during observations of the deck and has lower spatial resolution, it is an essential complement to the Pancam for estimating dust thickness and mineralogical composition. Estimates of dust loading from the Pancam and Mini-TES observations would be compared to estimates of solar cell power output using the two cells that are designed to be monitored for voltage and current values. When coupled with solar optical depth and sky brightness measurements from Pancam observations, the ensemble of data should provide quantitative estimates of dust loading and consequent power degradation.
 Basic physical properties of Martian soil-like materials can be estimated from interactions of the rover wheels with these materials during traverses and trenching activities (Tables 1, 2, and 4). This information will be used for characterizing and distinguishing among different surface units, and for determining stratigraphic relationships (especially as revealed by wheel trenching). In pursuit of these goals, rover wheels will compress, scratch, and dig into the surface, using methods expanded from previous experience with Sojourner [Rover Team, 1997a, 1997b; Moore et al., 1999] (Table 4). Wheel/soil interactions during rover traverses and dedicated soil mechanics maneuvers will be combined with supporting Hazcam, Pancam, and MI images (Table 3) to allow angles of internal friction, cohesion, particle size-frequency, and bulk density of surface materials to be estimated, with other payload elements contributing composition and mineralogic information.
Table 4. Summary of Mars Exploration Rover Soil Physical Properties Experimentsa
Note that APXS, MB, and Mini-TES can be applied for compositional and textural information on wheel tracks, wheel drag scars, and wheel trench walls (and tailings) as resources allow.
Fidelity of tread casts in Hazcam, Pancam, MI images
Clues to the presence of otherwise unresolvable clay- and/or silt-sized particles in soil-like materials
Crispness of track margins in Hazcam, Pancam, MI images
Qualitative assessment of soil cohesion, presence of otherwise unresolvable clay- and/or silt-sized particles in soil-like materials
Cracks in surface layer near tracks in Hazcam, Pancam images
Presence of very weak surface crust, not recognizable unless driven over
Estimate of soil-bearing strength, compaction state, compressibility
Albedo and spectral differences between drag scar and surrounding ground in Hazcam, Pancam images
Presence of spectrally distinct veneer too thin to recognize otherwise (e.g., an airfall coating)
Textural differences between soil components inside and outside drag scar in Hazcam, Pancam, MI images
Presence of texturally distinct veneer too thin to recognize otherwise (e.g., a particle lag deposit)
Steepness of sides of drag scar tailings in Hazcam, Pancam, MI images
Angle of repose
Cracks in surface layer near drag scar in Hazcam, Pancam images
Presence of weak near-surface crust, not recognizable (even if driven over - too deep) unless plowed
Trenching wheel motor currents, suspension orientation telemetry
Cohesion, angle of internal friction, estimates of bulk density, detection of major strength horizons (i.e., soil horizons such as crusts or buried obstructions).
Details of trench walls in Hazcam, Pancam, MI images
Local stratigraphy based on textural or spectral contrasts. Angle of repose of trench walls. MI only: Particle-size frequency of sand-sized and larger grains, superior textural assessments of walls.
 As noted in a previous section of this paper, the level of effort and downlink expended on soil mechanics experiments will depend on sol-to-sol operational constraints. Much will be learned from rover maneuvers and observations carried out for other purposes. For example, a single rear Hazcam frame of any wheel track can be examined for clues to the presence of otherwise unresolvable clay- or silt-sized particles by evaluating the fidelity of the wheel tread casts in soil-like material. Stereo Hazcam images of the same wheel track will allow more quantitative assessments of sinkage and calculation of other physical parameters (see following section). With a slightly larger operational investment, the MI can be placed on soil-like unit whenever front Hazcam data are available in order to determine particle size-frequency and shape (of sand-sized and larger grains) at stops along rover traverses. Such observations will be mapped on Pancam and Navcam mosaics to help define distinct surface material units and identify candidate sites for dedicated soil mechanics maneuvers involving wheel-dragging and wheel-trenching.
 Our approach for the analysis of Martian soil-like materials is guided by principles of terrestrial soil mechanics. Terrestrial soils are complex substances that can exhibit brittle, elastic, or plastic deformation under different stress regimes. Mechanical behavior of soil-like materials can vary depending on deformation rate, water content, mineralogy (especially clay content and type), climate effects, and stress history. Soils derive their strengths through particle friction, and a cohesion component is often present as well (e.g., from cementation, chemical bonding, or electrostatic attraction). Strength properties of soils can be described in a number of ways, but the most relevant characterization for our purposes, involving deformation at very low normal stresses, is the Mohr-Coulomb relation:
where τ is shear stress along the failure plane, σ′ is effective normal stress on the failure plane, ϕ′ is the effective stress friction angle, and c′ is soil cohesion. When the pore fluid pressure within the intergranular voids is zero, the effective normal stress, σ′, and the total normal stress, σ, are the same. This is the case for Martian soil-like materials, which are expected to be dry at the surface (by terrestrial standards, at least within the uppermost 12 cm). Values of ϕ and c for a soil can be determined by performing several controlled shearing failure experiments under different normal stresses. With σ and τ values plotted as (x, y) pairs, the experiment results scatter along a straight line, with tanϕ and c represented by the slope and y-intercept, respectively. The Mohr-Coulomb relation has been employed extensively and reliably by geotechnical engineers for many decades to describe shear strength behavior of a soil for a range of normal stress conditions. Several factors influence the angle of internal friction, ϕ, of soil-like materials, including void ratio, grain angularity, and sorting.
 During the Pathfinder mission, mechanical properties of soil-like material were derived mainly from analysis of Sojourner wheel tracks and wheel-trenching sites. As discussed above, images of soil-like material disturbances were used for estimating whether significant amounts of very fine particles were present, whereas friction expressed from motor currents measured during wheel trenching was used for quantitative estimates of soil ϕ and c values. The main technique for determining ϕ and c involved holding five wheels fixed while rotating the sixth, then plotting matched pairs of shear and normal stress values recorded during wheel-trenching. A similar trenching technique will be used by the MER rovers to estimate ϕ and c values, and to provide access to subsurface materials for the Athena instrument suite. The front wheels on each rover will be used for wheel-drag and wheel-trenching experiments because they allow the most direct access to disturbed soil-like material for the MI and other instruments on the IDD, and they require smaller drives to expose disturbed soil-like material to view by Pancam and Mini-TES. Also, only front wheels carry the temperature sensors necessary for interpreting the wheel motor performance against friction with soil-like material produced during digging.
 Surface materials distinguished by their mechanical properties will be correlated with surface morphologies seen in images and Mini-TES data (including particle size-frequency characteristics, bedding, and other textural and stratigraphic clues exposed by trenches into soil and aeolian bedforms), MOC coverage, surface units in THEMIS day/night image pairs, as well as other data, and will provide “ground truth” for improved modeling of orbital data sets. Physical properties of different surface units sampled along rover traverses will be compared and contrasted with surface materials characterized at the Viking and Pathfinder sites [e.g., Moore et al., 1987, 1999; Rover Team, 1997b] (Table 1), although this task will be complicated by differences in hardware, methods, and instrument calibration.
7.2. Terramechanical Studies
 Practical mathematical formulations developed to describe terrestrial vehicle/soil interactions (the study of terramechanics) combine theoretical and empirical approaches. Soil parameters oriented toward vehicle trafficability have been related to wheel sinkage, wheel dimensions, and wheel vertical load, as well as the slip of powered wheels (see summaries by Bekker [1956, 1960, 1969] and Wong ). These formulations also account for sinkage contributions from several wheels following each other in the same track. The MER case is complicated somewhat by slight inward offsets of the middle wheel tracks compared with the tracks created by the front and rear wheels.
 If a flat plate is pressed into a soil with a range of pressures p, vertical displacements z can be described with a function of the form (Bernstein  as cited by Bekker [1969, p. 39])
where k* is Bernstein's modulus and n depends on the characteristics of the particular soil. Soil-like materials that strengthen with depth have n > 1; soil-like materials that deform more easily at greater depths have n < 1. If two flat plates of different width (b1 ≠ b2) are used to penetrate the same soil to a series of depths under a range of pressures, the problem is closed sufficiently to resolve k* further into
where kc and kϕ are referred to as the cohesive and frictional moduli of soil deformation, respectively, and are soil specific [Bekker, 1956, 1960, 1969]. Measuring n, kc, and kϕ requires the use of a “bevameter” in which plates of various dimensions are loaded and soil sinkage amounts are measured. Further analysis using kc and kϕ normally can lead to estimates of c and ϕ, but this probably is unlikely for MER because the required bevameter data cannot be acquired with the rover vehicle hardware. The Mössbauer contact plate sensor, designed to prevent excessive sinkage of the Mössbauer instrument head into soil-like material, stops IDD movement at approximately 4 N of force, equivalent to the sensor head pressing into the soil-like material with about 1100 Pa of pressure. The resulting imprint may have morphological implications in the weakest of materials, but this single (z, p) data point is insufficient to determine n, k*, kc, and kϕ. The single (z, p) data pair obtained from a Mössbauer sensor plate imprint might allow k* to be broadly estimated if an assumption of n can made from other observed or measured, or inferred soil-like material properties.
 Wheel imprints will be much deeper than Mössbauer sensor plate imprints into a given soil-like material, be more available for sinkage measurements from stereo images, and provide an alternative method for estimating k* if, again, a value of n can be assumed on the basis of other observed, measured, or inferred soil-like material properties. For a single rigid rolling wheel, the relationship between vertical wheel load W, wheel width b, wheel diameter D, the exponent of soil deformation n, the Bernstein modulus k*, and the wheel sinkage z0 at wheel bottom-dead-center is given by [Richter and Hamacher, 1999]
Sinkage of a second rigid wheel under load W following behind in the same track (applicable, approximately, to the center wheels of the MER rovers) is described by
with z2 = additional load-sinkage of the trailing wheel, beyond z0 created by the first wheel. Total sinkage of the trailing wheel is thus z0 + z2. A third trailing wheel, corresponding to the rear wheel of the 6-wheeled MER vehicles, will create load-sinkage with wheel load W according to
where z3 is the sinkage beyond that of the first two wheels, and z0 + z2 + z3 is the total load-sinkage of a rear wheel relative to undisturbed soil-like material surface. All rover wheels are powered, so sinkage resulting from slip, zs, must be accounted for, described by
where i is wheel slip and hb height of wheel tread grousers, if present. Only the total, multiwheel track depth can be observed and measured in images of wheel tracks behind the rover. A function relating n and k* can be obtained by iteration of equations (4)(5)–(6). An assumed pair of (n, k*) values can be evaluated by comparison with predictions of these variables from calculating first z0, then z2, and finally z3 successively from equations (4)–(6), using wheel diameter D, width b, and effective vertical wheel load W from suspension/chassis telemetry. A reasonable value for wheel slip i must also be assumed (practical for traverses over principally level terrain) to allow sinkage resulting from wheel slip to be removed from track depth measured from images, isolating depth resulting solely from load-sinkage. Iterations can be continued until an optimized function relating n and k* is obtained (but not specific values of either parameter) consistent with measured total track depth z0 + z2 + z3 measured from analysis of stereo images. A specific estimate of k* can then be made with an assumed value of n.
 Whether ϕ can be estimated from sinkage measurements or determined from wheel-trenching as discussed previously, soil bearing strength can be calculated from wheel tracks by adapting the relationship for critical load per unit length Wc from bearing capacity theory [Terzaghi, 1948],
with γs soil weight density (in Nm−3), b the width of the loading surface (e.g., wheel), q surcharge acting onto the soil surface (usually zero), c soil cohesion and Nγ, Nq, and Nc bearing capacity factors which only depend on the soil friction angle ϕ. A distinction is made for the functions Nγ(ϕ), Nq(ϕ), and Nc(ϕ) depending on whether the soil can be considered “dense” (i.e., showing little compressibility) or “loose” (exhibiting significant compressibility) but the relationship with ϕ is classically available from bearing capacity theory in both cases.
 Wheel sinkage described in equations (4)–(6) will be calibrated specifically for the MER wheel, because rather small wheel diameters require corrective factors to improve the accuracy of results [Richter and Hamacher, 1999]. This calibration will also be useful for rover systems analysis described in the following section.
8. Terrain Property Estimation Using Rover Dynamics During Traverses
 During traverses the rovers will be sampling at 8 Hz information that can be used to retrieve the topography of terrain, along with the mechanical properties of soil-like materials and rocks encountered along the way. The telemetry returned will include, if commanded, time series that show the rover tilt vector, the rocker and rocker-bogie angles, wheel currents and encoder values (i.e., number of wheel turns), and wheel azimuths (e.g., Figures 11a and 11b). We will utilize these parameters acquired along traverses to model the entire rover as a mechanical system that is responding to terrain topography and material properties through its powered wheels and suspension system. The intent is to model the overall system in such a way that the terrain topography and material properties can be retrieved from the traverse data, using both forward modeling approaches and non-linear inversions.
 Specifically, telemetry data will be analyzed to estimate terrain slope angle distributions. Cohesion and angle of internal friction values will be retrieved for portions of the traverses during which wheel slippage occurs with soil-like materials [e.g., Iagnemma et al., 2002]. Imaging data and other physical properties experiments described in this paper (e.g., wheel track analyses and trenching activities) will be incorporated into the analyses to refine estimates of parameters extracted along the traverses (including slopes of drift faces, dunes, and ripples). The traverse terrain parameter retrievals will be accomplished using a finite element and numerical integration approaches for modeling masses associated with various rover elements, rocker and rocker-bogie suspension system dynamics, and wheels and their mechanical characteristics, including torques associated with wheel motors (Figure 11). Further, the modeling will be used to help understand rover performance for the various terrains over which the vehicles traverse and perhaps to evaluate possible traverses before actually attempting them on Mars.
9. Synergistic Analysis to Address Outstanding Questions
 Full realization of the scientific and technical potential of experiments described in this paper will occur only by integrating observations from the entire ensemble of Athena instruments (Tables 1–4), and orbital observations. Answering the outstanding scientific and technical questions summarized in Tables 1 and 2 requires that such an approach be followed. For example, the Rock Abrasion Tool will provide direct information on rock mechanical properties through analysis of the motor currents and penetration rates as the system abrades away the outer 5 mm of the rock surface [Gorevan et al., 2003]. The MI will provide direct measurements on grain size distributions for soil-like materials and rocks [Herkenhoff et al., 2003]. Mini-TES observations will be used to infer surface thermal inertias and thus direct information on the degree of induration of soil-like materials, together with grain size distribution information and mineralogy [Christensen et al., 2003]. Pancam color and photometric surveys will provide mineralogical and textural information for soil-like materials and rocks [Bell et al., 2003]. The magnetic properties experiments [Madsen et al., 2003] and MB Spectrometer and Alpha Particle X-Ray Spectrometer Experiments [Klingelhöfer et al., 2003] will provide important information on the mineralogy and composition of the various soil-like materials traversed and trenched, along with critical information about rocks, both before and after removal of coatings using the RAT. Thus all available data will be used to help address the major scientific and technical questions posed in this paper, using teaming arrangements that optimize synergistic analyses.
 Finally, a major challenge will be to take the entire ensemble of MER-related observations and analyses to understand orbital observations in much more detail than otherwise possible. That is, the MER data will be used to provide ground truth to calibrate what is seen from orbit and to extrapolate to regions not covered by landed observations. For example, Figures 1 and 2 show the albedo and thermal inertia characteristics of the Meridiani Planum, and Gusev Crater MER landing sites, in addition to the Viking Lander 1, Pathfinder, and the Elysium Planitia and Isidis Planitia MER backup sites. The Meridiani Planum site is located on the top stratum of an exhumed layered sequence [Christensen et al., 2001b; Hynek et al., 2002; Arvidson et al., 2003]. The unit exposes hematite, perhaps in dark dunes, and underlying deposits are widely exposed elsewhere in the region. The terrain to be observed on Meridiani Planum from MER certainly will be different than observed at the Viking Lander 1 or Pathfinder sites, given the morphology and mineralogy inferred from analyses of orbital data and the low-albedo and thermal inertia values as compared to these other two sites (Figure 1). Gusev Crater may be the dustiest landed site thus far (Figure 2). In addition, extensive duricrust might be expected at the Gusev Crater site (Figure 2). Relating the MER observations to inferences from orbital measurements will provide the critical links to be able to explain with improved confidence than now the nature of surface materials inferred from morphologic, elemental, and mineralogic mapping from orbital platforms.
 We thank the Mars Exploration Rover Project for support and Lutz Richter also thanks the German Space Agency. We thank Don Bickler, JPL, for insightful discussions and Laurence Soderblom, Kim Deal, and an anonymous reviewer for their thoughtful comments and suggestions. Megan Murphy provided critical technical assistance in preparing this manuscript.