P.H. Smith, M.G. Tomasko, D. Britt, D.G. Crowe, R. Reid (Lunar and Planetary Lab, University of Arizona, Tucson, AZ 85721)

H.U. Keller, N. Thomas (Max Planck Institute for Aeronomy, Lindau-Katlenberg, Germany)

F. Gliem, P. Rueffer (Technical University of Braunschweig, Braunschweig, Germany)

R. Sullivan, R. Greeley (Dept. of Geology, Arizona State Univ., Tempe, AZ)

J.M. Knudsen, M.B. Madsen, H.P. Gunnlaugsson, S.F. Hviid, W. Goetz (Niels Bohr Inst. for Astronomy, Copenhagen, Denmark)

L.A. Soderblom, L. Gaddis, R. Kirk (USGS, Flagstaff, AZ)

Abstract

The Imager for Mars Pathfinder (IMP)­a stereo, multi-spectral camera­ is described in terms of its capabilities for studying the Martian environment. The camera's two eyes, separated by 15.0 cm, provide the camera with range-finding ability. Each eye illuminates half of a single CCD detector with a FOV of 14.4x14.0° and has 12 selectable filters. The f/18 optics have a large depth of field and no focussing mechanism is required; a mechanical shutter is avoided by using the frame transfer capability of the 512 x 512 CCD. The resolving power of the camera, 0.98 mrad/pixel, is approximately the same as the Viking Lander cameras; however, the signal-to-noise ratio for IMP greatly exceeds Viking approaching 350. This feature along with the stable calibration of the filters between 440 and 1000 nm distinguishes IMP from Viking. Specially designed targets are positioned on the Lander; they provide information on the magnetic properties of wind-blown dust, measure the wind vectors, and provide radiometric standard reflectors for calibration. Also, eight low-transmission filters are included for imaging the Sun directly at multiple wavelengths giving IMP the ability to measure dust opacity and potentially the water vapor content. Several experiments beyond the requisite color panorama are described in detail: contour mapping of the local terrain, multi-spectral imaging of the surrounding rock and soil to study local mineralogy, viewing of three wind socks, measuring atmospheric opacity and water vapor content, and estimating the magnetic properties of wind blown dust. This paper is intended to serve as a guide to understanding the scientific integrity of the IMP data that will be returned from Mars starting on July 4, 1997.

Introduction

The Imager for Mars Pathfinder (IMP) was conceived in January 1993 as a response to NASA AO No. OSSA-03-92 which outlined the requirements for a camera to support what was then called the MESUR (Mars Environmental Survey) Mission. MESUR was expected to be composed of a network of landed stations on the surface of Mars with the Pathfinder lander being the engineering prototype. With the replacement of MESUR by the Mars Surveyor Missions, Mars Pathfinder has become the first of a series of independent scientific payloads to explore the Martian surface. IMP is a key element of this program.

The heritage for IMP derives from the Descent Imager/Spectral Radiometer (DISR: Tomasko et al., 1996), one of the primary instruments aboard the Huygens probe that will descend through TitanÕs atmosphere by parachute in 2004 as part of the CASSINI Mission to the Saturn system. The heart of DISR is a Focal Plane Array (FPA) and readout electronics package supplied by the German co-investigators (lead by H. U. Keller) at the Max Planck Institute of Aeronomy (MPAe); it is a 512 square CCD array that rapidly transfers the image charge to the storage section in 0.5 msec (no mechanical shutter is required). The chip is then slowly read out through a 12-bit ADC over the next 2 sec. This CCD has been developed by LORAL in Huntington Beach, CA where 8 different lots have been processed to perfect the design. Other components were available from DISR at an advanced stage of maturity including imaging lenses and a power supply board, both designed by Lockheed Martin Astronautics (LMA). The following pages describe the camera system and its calibrated properties. Then each experiment (panoramic imaging, stereo, mineralogy, atmosphere, wind, and magnetic properties) is individually discussed in terms of science goals, operations and data analysis techniques.

IMP Description

The Viking Lander cameras set a high standard for excellence after landing on the surface of Mars in July, 1976. With a FPA consisting of 12 individual silicon detectors and a spin-scan rastering system, color, stereo pictures were returned from 2 locations. Now, 20 years later, the second camera system to land on Mars is lighter, more capable, and vastly cheaper; yet the final pictures will look qualitatively similar to those of Viking. The major difference is that a different location on Mars will be the subject for these pictures.

The IMP was designed for a resolution of 1 mrad/pixel, about the same as Viking (color: 2.1 mrad/pixel, monochrome high-res: 0.7 mrad/pixel). The design specifications are summarized in a tabular form in Table 1. The FPA, supplied by MPAe, retained the DISR specifications to meet cost, schedule and reliability constraints, so that the 23 µm pixel spacing and the 1 mrad/pixel resolution requirement imposed by the AO determined the focal length of the camera as 23 mm. No attempt was made to improve on this resolution because of the low data rates that were expected during the mission. Our decision to include stereo plus the requirement for low mass led us to split a single detector into left and right image areas rather than deal with the extra complexity of two separate focal planes. Therefore, there are two identical 248x256 pixel sub-arrays for each eye separated by a 12 pixel "dead zone." Each FOV is 14.4°(H)x14.0°(V). We pushed the eye separation as far apart as the volume allocation would permit to get a 15.0 cm separation between the eyes, the light path is folded by two sets of mirrors to bring the light to the FPA.

Early engineering tradeoffs balanced the conflicting requirements of low mass and of high strength to survive the 100 g landing shock expected through the use of a challenging airbag-assisted bounce onto the surface of Mars. Titanium was used where possible because of its large strength-to-mass ratio. The gimbal is held with duplex bearing pairs that are dry lubricated to maintain functionality at low temperature. Motor/gearhead combinations, supplied by the American Technology Consortium (ATC), are positioned along both the azimuth and elevation axes. The motors are 4-pole steppers with an 81:1 gearhead ratio, giving a step size of 0.553° and driven at 10 steps/sec; the step count from a hard stop is maintained to determine position.

A mast, supplied by AEC Able Corp., of the open lattice type used by magnetometers is held in its stowed position during cruise by a pyro-activated pin puller. At the end of the first sol the pyro is fired and the camera pops up 80 cm above its stowed location, about 1.75 m above the surface. From this vantage the horizon for a featureless sphere is 3.4 km away. The flight camera deployed on its mast is shown in Figure 1.

To accommodate the large elevation range, a cylindrical shape was judged to be the most convenient with the elevation axis running through the center of the cylinder. Putting the optic axis 12 mm above the center line makes downward pointing easier. A dark, dust tight position is achieved by looking straight down into the yoke; small brushes protect the window housing from dust intrusion. Additional dimensions and an interior diagram of the camera are shown in Figure 2. The primary area for stereo imaging, supporting Rover operations with contour maps and distance information, is from 2-10 m from the camera. Therefore, a toe-in of the two fold mirrors of 18.5 mrad each gives complete overlap of the left and right eyes at 4 m distance. The keystone distortion caused by this toe-in is judged not to be significant.

An intensive effort was made to reduce light scattering inside the camera head. A series of baffles, darkened with Martin Black, are used to reduce the stray light to acceptable levels. A knife edge baffle reduces crosstalk between the left and right eyes to undetectable levels. The outer windows are recessed in their housings until they nearly touch the fold mirrors so that the housings act as sunshades to shade the windows from direct sunlight which can scatter off of dust particles.

The lenses are modified Cooke triplets stopped down to f/18. The PSF for the lenses has a half width about the size of a single pixel; lens aberration and distortion are unmeasurable. An example of a color picture taken with the IMP is shown in Figure 3, the subject is a model of the Rover positioned in the LPL Mars Garden at the University of Arizona. Note the green olivine nodules that are clearly visible in the nearby rocks.

There are three electronic boards housed in the lander chassis where they are always maintained above -50° C and are connected via the VME bus link to the onboard computer. A copy of the DISR power supply creates all the voltages for the CCD system from the Lander 28V power bus. The second board is the MPAe board that sends the clock pulses to the CCD and receives the analog signals back from the pre-amp board and converts them to digital signals to be stored in the frame buffer chip on the third board. The frame buffer board is connected to the VME backplane and is controlled by a Field Programmable Gate Array functioning as a state machine for command decoding. Another function for the frame buffer board is phasing the steps and driving the three motors.

The camera system is controlled through a sequence of uplinked commands that are time tagged and stored in RAM until they are required. The basic modes follow the operational sequences. Upon landing the first operation is the release of the launch lock, a wax thermal actuator pinning the camera head during takeoff, cruise, and Entry, Descent and Landing; this restores the full range of all camera motions. The image command includes many optional parameters that control the exposure and processing. Everything from the exposure time to the amount and type of data compression are specified here and attached to the data set to be placed in the header. Sub-framing boundaries and pixel averaging parameters can also be specified. After processing, the packetized images are stored in the telemetry buffer.

Several types of data compression are included in the IMP software package. Lossless compression (1.3:1 to 2:1 depending on the scene entropy) using the Rice algorithm developed at JPL will be the workhorse for the IMP images as long as we have a data rate of several thousand bits per sec. For non-science or low data rate scenarios, a modified JPEG compressor using arithmetic coding developed at the Technical University of Braunschweig will normally be used at ratios from 6:1 to 24:1. It is both enhanced by Local Cosine Transform prior to the JPEG-specific Discrete Cosine Transform and made robust against data dropouts (Rueffer et al., 1995). Other methods include sub-framing the image, for instance most pictures of the Sun will be returned as 30x30 pixel blocks centered on the Sun image. Row and column averages will be used for sky images, this gives the gradient and the edges of cloud features, but not the high resolution of an image. Pixel averaging (m x n pixels) can be used were full resolution is not needed. All these methods can be used in combinations for highest compression.

An uncompressed IMP frame contains 248 x 256 x 12 bits, or 762 Kb. About 100 frames are combined into a monochromic panorama which brings the data volume to 76 Mb; instrument and packet headers add another 5% for a total of 80 Mb. A typical day on Mars allows the return of 28 Mb of IMP data. Therefore, compression is an indespensible tool for the early mission when decisions on Rover traverses and Alpha-Proton-X-ray Spectrometric (APXS) sample sites must be made quickly. A 6:1 compressed panorama gives 13 Mb and is easily returned the first sol. Since the IMP mast deploys at the end of sol 1, multi-spectral panoramas will be gathered and stored as an insurance policy against camera failure; 550 Mb of RAM is available for this purpose.

Laboratory Calibration

The IMP has been characterized and calibrated with respect to absolute responsivity, spectral response, image quality, flat fielding, stray light, and pointing accuracy. This calibration data for the flight model will be used in combination with experience to be gained with the prototype model during the year before Mars landing, in order to further characterize the ability to perform scientific experiments in situ. The primary calibration took place during November and December 1995, a second opportunity in May and June 1996 has been helpful to fill gaps in the data set. The data set comprises about 10 GB on optical disk storage (CD-ROM).

CCD Properties

The CCD deviates from linearity by less than 1% and quantum efficiencies approach 50% at the peak falling rapidly to zero near 400 nm and 1050 nm (see Figure 4). The electronics readout is about 15 electrons so that with a gain of 30 electrons/DN the system noise is typically 1 DN at cold temperatures, where DN is the number of counts from the analog-to-digital converter (ADC). With a 12-bit ADC, the shot noise will limit the SNR to values of 350 per pixel. The anti-blooming gate initiates an exposure and captures any overflow charge while the frame transfer ends an exposure after a pre-set exposure time.

The dark current generated when the CCD is above -20° C is important to the reductions of the images. A dark frame consists of a combination of a bias offset (8.7 DN), the noise generated in the readout process, and the dark current gathered during an exposure. Ambient temperatures on Mars during the day will cause the temperature of the CCD to rise above 10ûC in the afternoons. The average total dark current from the CCD as a function of temperature for a one second integration time is shown in Figure 5.

The dark current from a "dark strip" of seven columns of detectors next to the left eye, shielded from light with an Al overcoat, is used to scale dark current corrections. Periodic dark frames are obtained by placing the camera in a stowed position, selecting a solar filter and taking a full frame. These dark frames are then scaled to match the dark current of a later image using a row-by-row ratio of dark strips. The onboard algorithm has been modified to compensate for a small light leak into the ÒdarkÓ strip.

Flat Fielding

The extent to which responsivity variations across the detector CCD array can be corrected ("flat fielded") in a single image is limited by the noise in the image, and by the accuracy with which the CCD spatial response variation is known. The flatness of the integrating sphere target is only known to about 1%, which is the dominant limitation for removing IMP response variations. Earth-based flat fielding from laboratory measurements currently achieves this level of performance. Possible changes in the IMP spatial response pattern, particularly due to dust on the entrance windows, may require subsequent calibration from images of the Martian sky.

A typical image of a flat field shows a cross-hatch pattern inherent in the manufacture of the detector chip superposed on a cos4 fall off that is characteristic of imaging systems. The high spatial frequency pattern has an amplitude of a few percent and is not noticable in the images while the center-to-corner fall off approaches 10%. The response flatness has been attained in ground data reduction for each filter, several filters have been averaged to provide a flat field correction capability on board the IMP. The challenge for on-board flat fielding precision is created by the fact that only one 8-bit correction table can fit in the non-volatile memory allocation. Image data compression algorithms depend upon pixel-to-pixel correlations in the image. High spatial frequency responsivity variations in the IMP would introduce artifacts which make compression less efficient. The response pattern over the FOV of each IMP eye is a function of wavelength, making it impossible to fully correct to an optimally flat field at every wavelength using a single table.

Many choices for creating a flat field table for the flight software were tested for their quantitative performance. The choice which has been made is to correct using an average of the tables for filters 5 through 11 excluding the diopter filter in the right eye. Filter 0 (440 nm) was omitted because the response pattern at that wavelength is almost anti-correlated with the longer wavelength response patterns. At the short 440 nm wavelength, the response variations within each pixel become part of the pattern, while at longer wavelengths increased diffraction blur masks this part of the pattern. The resulting flight software tables will only be used for the filters from which the average was taken. Proper flat fielding corrections will be made during ground processing.

Absolute Responsivity

The absolute responsivity for the geology filters at a typical CCD operating temperature of -20ûC is shown in Table 2 along with the central wavelength and bandwidth of each filter. The central 100 pixels of each eye have been averaged while looking at an absolutely calibrated target at temperatures of +25°, 0°, -20°, -60°, and -110°C. A comparison of two absolute calibration sources was used: a Labsphere-calibrated integrating sphere, and a JPL standard lamp; the responsivities agreed to within 10%. Figure 6 illustrates the temperature dependence of the center wavelength and bandwidth of one of the solar filters. Figure 7 is a computed set of Mars exposure times for a surface scene under the following conditions: Mars distance from Sun 1.52 AU, dust optical depth 0.3, Solar Zenith Angle 45°. The effects of the lower Mars surface brightness in the blue, and the decreasing CCD quantum efficiency in the infrared, cause longer exposure times at the extremes despite wider filter bandwiths. Table 3 provides the solar filter absolute responsivities at -20°C. Typical solar exposure times will be a few msec near noon and increase rapidly when the Sun nears the horizon. The geology and solar filter relative spectral responses are shown in Figures 8 and 9. The temperature dependence will be important in the narrow solar filters, which are to be used for detecting dust opacity and water absorption bands.

Image Quality

The Cooke triplet lenses used in the IMP provide unmeasurable geometric distortion (below 0.05 pixels) across the field of view. The Line Spread Functions (LSF) and Modulation Transfer Functions (MTF) also do not vary significantly over the field of view. Models predict blue MTFs that agree within 5% of the measured values. The LSF widens in the tails more with wavelength than the models can explain. This produces a measured MTF at 965 nm which is only 60% of the theoretical optical value. This excess broadening is attributed to photoelectron crosstalk between electron wells, due to the increased transparency of silicon to photons at longer wavelengths.

The MTFs all illustrate that the IMP imaging system, which is band-limited by the active area of a pixel, exhibits significant aliasing at all wavelengths. The combination of diffraction blur and electron crosstalk makes aliasing less significant at longer wavelengths. This is a primary reason for including the 965 nm stereo pair in the filter selections. The 965 nm wavelength will be used for the highest precision stereo ranging measurements because it causes the least systematic error at sharp edges.

Stray Light

The IMP stray light calibration is intended to give as much information as possible to determine the accuracy of quantitative image measurements, and to provide data which may be useful in correcting stray light contributions. A map of the stray light due to an intense source subtending the same angular size as the solar disk at Mars (6 mrad) has been made for angles up to 50û off-axis. A series of increasing overexposures allows quantification of stray light down to about 3 ppm of the source peak. The stray light which takes the form of out-of-focus dust particle images monotonically decreases with angle away from the source, and has fallen below detectability by about 35° off-axis.

The IMP images over a field of view about 14° square, but light from a larger full angle (about 80°) enters the camera window directly. The amount of light appearing within the FOV due to illumination outside the field has been measured for the IMP. Integrated stray light due to overfilling the field of view increases with scene subtense half angle at the entrance window. Completely overfilling the field results in integrated stray light of 5.03% in the left eye, and 2.12% in the right eye. The forward scattered stray light was integrated by using a large white target with a small black area in the center. As the target approached the camera from the distance at which it just filled the field of view, the size of the black area increased, decreasing the amount of forward scattered light. The integrated forward scattered light at the center of the field of view is about 3% in each eye.

Ghost images of bright objects appear in the IMP due to the varying wedge angles between filter faces. During calibration, a set of ghost images was made by overexposing a 6 mrad diameter source against a dark background. The ghosts tended to be limited to one or two faint reflections, with the brightest near 2% of the source intensity.

Pointing Accuracy

The IMP has azimuth and elevation axes driven by 4-pole stepper motors with planetary gearhead reducers (81.3:1), a single step produces 0.553û of motion. The initial sun search performed with the IMP must locate the sun within 1û for pointing the high gain antenna at the earth. Science operations on Mars do not require higher pointing accuracy, only the ability to subframe objects within the field with confidence. The pointing data has resulted in a flight rule: when accurate pointing is desired, azimuth movements will be commanded with a history of at least 80 motor steps from the same sense of rotation to maximize backlash predictability by pre-loading the cable drag. The azimuth backlash is approximately 1° except near the extremes of the travel where it becomes less predictable due to the cable drag. Elevation backlash is smaller, and is a function of elevation angle. The pointing accuracy, when the flight rule is followed, has been demonstrated in integrated spacecraft system tests to be within 5 pixels in azimuth and 6 pixels in elevation.

Color Imaging

Color imaging is an important IMP task. The IMP makes color images in one eye from red, green , and blue filters. Stereo color images will be made using the RGB information from one eye in combination with RB information in the other eye to deduce RGB for that eye. This method frees one of the RGB stereo pair filters in one eye to create an IR stereo pair at 965 nm for reduced aliasing in stereo ranging measurements. The feasibility of synthesizing green in one eye from red and blue was established in a series of test images. The green channel which was created was compared to an actual green image for the test. A one sigma error of 6% is present in the green channel images made with this preliminary algorithm. That 6% error is reduced in its apparent effect to the human eye by the fact that it appears with correct red and blue channels, and a stereo mate image with correct RGB. The resulting stereo pair RGB image does not appear distinguishable from an RGB stereo pair image made with data rather than one predicted green channel. True color comes in two varieties on a Mars mission: Mars Color and Earth Color (the color Mars objects would have if transported to the earth). Either ÒtrueÓ color can be created in principle by using the absolute spectral calibration. A relative method will also be available using images of the color targets on the lander and rover. The atmosphere of Mars can produce a family of ÒtrueÓ Mars colors due to the variation in the amount of dust reddening of the solar illumination. Quantitative spectra and true earth color anchor these ambiguities in a constant color context.

The Panoramic Images

The instrument described in the previous sections provides a powerful tool to determine the topography and geology of the landing site, the photometric characteristics of the local surface, the homogeneity of the landing site over a wide range of spatial scales, and the appearance of sub-surface layers. The IMP will also search for long time scale variations produced by atmospheric effects, for instance, aeolian modification of dune structures. In addition, it will observe effects resulting from the landing of the spacecraft and of the motion of the Rover and hence place constraints on the physical properties of the surface regolith. Finally, the imaging system will be used to identify targets for the Rover and APXS experiments.

Before discussing the science goals a brief review of the surface operations is helpful. Pathfinder lands at 3 am LST on July 4, 1997 and retracts its airbags, positions the solar panels leveling the lander and prepares for the scientific mission. Shortly after sunrise, a Sun search algorithm begins a pre-programmed search along the horizon for the solar disk using one of the solar filters. When the search is successfully completed then the vector for the Earth is computed and instructions are sent to the High Gain Antenna to point it toward the Earth and begin high gain communications directly to Earth. Later in the morning IMP executes a panorama sequence in true color to image the Rover on its solar panel against the Martian landscape, this is called the Mission Success Pan although it is not a complete panorama. About mid-day the IMP will be commanded to provide a full panorama in the red filter. The last activity of sol 1 is the deployment of the mast.

Before deployment on sol 1 the highest resolution images of the nearest soil not obscured by airbags are planned using all the geologic filters and low compression. Inhomogeneity at millimeter scales and the potential to see beneath the weathered surface layers because of Lander disturbances make this set of images a high scientific priority. Many other images will be taken the first day and stored as an insurance policy against camera failure during the deployment operation which is rather violent and against potential freezing over the first night. Some of these panoramic images will be used as part of a vertical stereo pair with a similar pan taken after deployment. The landing process may have broken or crushed rocks on the surface. Given that the selected landing site is in a flood channel, the interiors of these rocks may reveal evidence of sedimentary formation. On the other hand, if the rock is volcanic in origin, the degree of crystallinity will provide information on the formation process. Furthermore, a detailed inspection of broken material at the highest resolution is needed to look at material unmodified by weathering processes. Broken rocks, uncontaminated by the pervasive dust environment, would be extremely interesting targets for the APXS.

For a perfectly smooth surface, with the "eyes" of the IMP 1.75 m above the ground, the horizon would be 3.4 km away at a resolution of 3.4 m/pixel. From many places within the present landing ellipse, high ground should be visible in the far distance. The position of the site with respect to local ridges, slopes, and channels will be important in the interpretation of the data at the site because the terrain will influence local winds and will have influenced deposition of water-borne materials. The local topography will also strongly influence the selection of the path to be taken by the Rover.

The first three-color panorama will provide an initial assessment of the range of materials. IMP will be used to characterize the morphology of the local surface materials. At the Viking landing sites, 10-20% of the surface was covered with rocks in the size range 0.035-0.45 m (Moore and Jakosky, 1989). Three other types of surface material were identified­drift material, crusty/cloddy material, and blocky material (Christensen and Moore, 1992). Drift material, about 14% of the surface near the Viking 1 landing site, was powdery; Arvidson et al. (1989) compared it to kitchen flour. Blocky material has a much higher cohesion and is heterogeneous. It has been compared to dry, cloddy, garden soil. Crusty/cloddy material has an intermediate cohesive strength and appears to be similar to terrestrial soils that have formed stable surfaces to wind erosion. The IMP will be used to identify the distribution of these types of material and assess the cohesive strength of surface material by studying the height and steepness of local slopes. The IMP will be able to make further investigations of the texture, composition, and strength of the soil by following the movements of the Rover and imaging the surface layers disturbed by the wheels. The Rover will apply a known force which will compress the surface layer. Both the stereo capability of the IMP and the Rover cameras will provide accurate measurements of the depth of the tracks allowing us to estimate the tensile strength of the surface and, possibly, its porosity.

The Ares Vallis site has been specifically selected to offer a wide variety of surface materials and the investigation of their photometric properties under a wide range of illumination conditions is a major scientific objective. The higher photometric accuracy of the IMP will allow a much more detailed assessment of the variegation based on color, shape, and texture. The phase dependence of the reflectivity will also allow us to estimate the roughness of surface materials and will provide further constraints on the heterogeneity and composition of the material.

The post-deployment images will include areas near the spacecraft already imaged before the mast was deployed. This combination of low- and high-level images will, in effect, give a stereo pair with a distance of 80 cm between the "eyes". These data will therefore allow high accuracy reconstruction of the size and shapes of rocks in the distant field. This can be used to determine the rock size distribution which may give some indication of the magnitude of the flow which produced the flood channel.

At the Viking 1 landing site, small (<25 m diameter) craters were much less numerous than expected by extrapolation of the size-frequency distribution derived from in-orbit remote sensing (Binder et al., 1977). A regional erosion rate at this site of several meters over geological time was estimated by Arvidson et al. (1979) which would be insufficient to explain this discrepancy. The absence of small craters may be evidence that the atmosphere was previously more dense that at present (Vasavada, et al., 1993). Observations of craters by the IMP will provide further constraints. Wind streaks extending from rocks are prominent features at both Viking lander sites (Sagan et al., 1977). These features are remarkably parallel and have been correlated to the obstacle diameter; they should be easily detectable at the Ares Vallis landing site. The streaks are interpreted as wind shadow areas and therefore give the mean wind direction during the last major storm. Interestingly, two distinctly different directions for the streaks were identified at the Viking 1 site. Sagan et al. suggested that the lengths of trajectories were important in determining the direction of the streaks because of the influence of Coriolis forces. IMP measurements should provide a further test of this theory.

The color capability of the IMP might also be used to search for evidence of aeolian sorting on the surface. Small particles may be stripped from dust layers by winds and deposited nearby. The difference in the scattering properties will lead to measurable changes in the surface photometry and will be particularly evident in the ratio of red and blue images of the surface. The IMP will therefore be able to identify relative changes in the surface particle size distribution. The spatial resolution of the IMP close to the spacecraft may be sufficient to observe the windblown motion of individual grains. Greeley et al. (1992) point out that the impacts of fine grains on larger ones may push grains downwind in a process known as "impact creep". The typical size range of grains affected by this process on Earth is around 3 mm and therefore resolvable by the IMP camera.

The large dynamic range of the IMP allows a sensitive search for time variability of the surface features. At the extremes of low temperature and high pressure condensation of atmospheric volatiles may occur. A dedicated search for evidence of surface frosts during morning twilight will be performed. Thin frost layers will be bright compared to the dusty surface, particularly at blue wavelengths, as was seen by Viking II. The monitoring of such time variable features will provide data on the atmospheric and surface temperatures and the thermal inertia of the surface.

Finally, the IMP will also be able to provide information on the performance of the landing system, although several of these aspects may be for curiosity only. Potential targets will include: a search for the location of the purple parachute and the back cover; a search for evidence of the grayish-colored residue from the solid rockets (mostly Al2O3); and investigation of places disturbed by the airbags as the Lander bounced across the surface.

IMP Stereo Data Processing

The fundamental IMP camera product is a stereo image pair (or an "image cube" of pairs obtained with the same geometry but with different filters). By design, such a pair can be viewed directly to give a three-dimensional impression of a portion of the IMP landing site. In addition, quantitative photogrammetric processing is planned that will lead to several useful derived data products. This processing will be performed by the U.S. Geological Survey on a commercial digital photogrammetric workstation (DPWS) after a limited amount of software modification.

Following collection of a series of stereo image (or cube) pairs by the IMP camera, a geometric control network will be established. Note that stereo information is obtained both from the 15 cm separation between eyes and also from the vertical separation of 80 cm achieved upon release of the pop-up mast. Manual or semi-automatic collection of the locations of matching features between paired images and between images from the same camera at different orientations will allow solution for both the location of the features in three dimensions and estimates of the camera orientations as revised from nominal values. These solutions in turn will permit mosaicking of image sets without offsets at the seams. In particular, measurement of locations on the local horizon will allow conversion from the spacecraft-centered coordinates to local geographic coordinates. This type of match-point collection and solution for camera orientations will need to be performed at intervals during the mission, both because the camera platform pointing is not perfectly repeatable and because it might indicate shifts in the lander's orientation. Match-point measurement of features on the rover (which has the advantage of known size and geometry) may be used to reconstruct its locations and orientations to higher accuracy than stereo imaging alone can provide. Finally, point measurements of distant landmarks in the landing site zone that are identifiable from orbit can, by a similar matching process, be used to locate the landing site precisely; this technique has been applied successfully in the Surveyor, Apollo, and Viking missions.

Once a control network has been established and camera orientations have been calculated, the next step is to match features in paired images on a point-by-point basis in order to determine the ranges to the features. In the current state of the art, the matching can be done automatically to an accuracy of ~0.2 pixel of parallax though only as averaged over patches on the order of 5-10 pixels minimum width. The corresponding precision of depth measurements deteriorates with range from the stereo baseline (see Figure 10 for ranging accuracy). On the other hand, the surface farther from the camera is viewed with increasing obliquity, so that the nominal precision in range (projected onto the ground) is everywhere about two pixels. Transverse resolution on nonhorizontal surfaces (e.g., facing surfaces of rocks) deteriorates less rapidly than the range resolution, however, so that shape measurements for distant rocks will be less satisfying than those for near ones. To state the obvious, roughly half of every rock in the landing site will remain unobservable and unmodellable based on IMP data alone. Integration of the Sojourner Rover camera images with the IMP data into a three-dimensional model might be used to address this problem, but coverage will be limited; furthermore the problem is highly demanding even with current technology.

With control networks, corrected camera pointing data, and image-coordinate range maps available, an impressive variety of data products can be generated. These products can be classified in at least three independent ways: mode of distribution, coordinate system, and data content.

For the mode of distribution, the data are detected, transmitted, and processed digitally, but there is a real need for both digital and hard copy data. Archiving of the full date set and distribution of products suited for quantitative analysis are best done digitally. Hard copy of summary data sets has complementary uses, however. Vastly greater volumes of data can be presented in hardcopy form than on a computer screen, and hardcopies can be marked on. In either mode, image, spectral, and topographic (geometric) information can be combined through the use of color, overlaid topographic contours, and so on. Both digital and hardcopy products can be divided further into ephemeral (on-line files, photographic prints) and more permanent (CD-ROMs, offset prints) categories.

For the coordinate system, as noted, the basic data are image pairs. With a control network, images from each eye (and filter and time) can be mosaicked together into a panoramic view from the lander's viewpoint. Experimentation will be done to determine the best lander-centered coordinate system to allow stereo viewing of left-eye and right-eye mosaics of this type because the IMP differs substantially from stereo camera systems on previous planetary landers. More radically, with range data any images can be reprojected or "orthorectified" from the camera coordinates to a hypothetical overhead view, that is, from a panoramic image to a map of the landing site. This process will be needed to merge spectral information from the filters that are not shared by the two cameras. Multispectral images merged in map coordinates could also be transformed back to the panoramic viewpoint of one of the cameras.

In terms of data content, a simple black-and-white panorama portraying the morphology of the landing site will be among the first data sets returned, with quasi-real-color stereo sets following soon after. As noted, stereo ranging is necessary for comparison of spectral channels that are not common to both cameras (see the next section for a description of the science goals addressed by the spectral filter set). Additional information can be obtained by a variety of multi-temporal observations, most importantly (a) detailed fits of photometric functions (using local surface orientations from the stereo models) to determine surface scattering behavior and hence to infer physical properties, and (b) change detection over short periods (e.g., searches for ephemeral frosts) or longer periods (changes in morphology, changes in lander orientation).

These types of observations can be synergistic. For example, observations with multiple illumination angles will not only constrain the photometric function of surface regions, but can lead to higher spatial-resolution modeling of the surface shape. Two-dimensional photoclinometry, or shape-from-shading (e.g., Kirk 1987), can be used to recover geometric/topographic information at pixel scale, generating much more detail than stereo matching can. Such analysis relies on an accurate photometric function model, while images with multiple illumination conditions can be used to differentiate brightness differences due to shape and shading from those due to intrinsic albedo differences. A coarse shape model based on stereo matching is highly effective as a starting point for a photoclinometric solution (Giese et al., 1996).

Multispectral Imaging

Most of our multispectral views of Mars are blurred by the large distances between the observers and the Martian surface. The best groundbased visible and near-IR instruments produce image cubes with ground resolutions of 100Õs of km/pixel. Recent HST images reduce this value to about 20 km/pixel. This large footprint averages large areas of potentially dissimilar materials and homogenizes the apparent surface mineralogy. Little spectral diversity has been observed; there are apparently two major spectral units, the so-called bright and dark areas. The dark areas have a reflectance of about 15% in the red and near IR, while the light areas are approximately twice as bright in the same spectral region, the blue albedo is nearly constant over the planet except for regions covered by clouds or ice. Representative spectra of the two regions are shown in Soderblom (1992).

Both types of material show a strong absorption in the blue to UV (300-550 nm) indicating the presence of oxidized iron. The spectra rise in the visible and plateau at approximately 700 nm with a number of weak absorption bands and shoulders. The spectral contrast of these features is low, due either to low abundances of a divergent mineralogy, or to a masking of spectral contrast by poorly crystalline weathering products. In general, the spectra indicate the presence of both coarse and fine-grained iron oxides, weathered basalts, pyroxenes of varying compositions, and as-yet unidentified hydrated minerals. Less certain are the presence of carbonates, sulfates, carbonate- and sulfate-bearing minerals like scapolite, and clay silicate minerals like nontronite.

The Martian environment is extremely harsh to exposed minerals, but there is little hard data on the exact nature of the surface weathering processes. The identification of these processes and of the mineralogy of their end products is a major science goal. The weathering reactions probably involve water, oxygen, and carbon dioxide, supplemented by UV radiation, the diurnal temperature cycle (over 100°C at the Pathfinder site), impact heating processes, and the abrasion of windblown dust. As the search for evidence of life on Mars intensifies, evidence for hydrothermal alteration of the rocks is an important goal for all spectral studies.

IMP Spectral Operations

The multispectral image cube is consists of 12 filters which are summarized in Table 4 along with their expected diagnostic information. The band centers of the filters are clustered (profiles are shown in Figure 8) to optimize discrimination of the two most important mineral groups expected to be found at the Pathfinder landing site. The first objective is identify the crystalline ferric oxides, oxyhydroxides, the poorly crystalline or nanophase ferric oxides. The major crystalline ferric oxide phases include hematite, goethite, maghemite, magnetite, and lepidocrocite (Morris et al., 1985). The basaltic weathering product palagonite is an example of a material that has visible to near-IR spectral properties doninated by nanophase ferric oxides (Morris et al., 1989,1990). The spectrally diagnostic region for these minerals is primarily in the UV to 860 nm range using 8 of the filters. This allows for discrimination based on four spectral parameters (Bell and Morris, 1995): (1) the position of the ferric iron UV drop-off (diagnostic of goethite, maghemite, magnetite, and lepidocrocite); (2) the position of the charge transfer band near 860 nm (diagnostic of hematite, goethite, maghemite); (3) the presence of a charge transfer band near 650 nm; and (4) the presence of a band between 650 and 800 nm. The depth of features at these diagnostic bandpasses can provide information on the crystallinity of the target material. Shown in Figure 11 are the laboratory spectra of 3 Mars-analog materials convolved to IMP bandpasses. In each case the IMP filter set provides several diagnostic parameters to allow the discrimination and identification of the mineral.

The second major goal is to characterize the mineralogy of the unaltered crustal materials that are composed of ferrous-bearing silicates. These minerals have absorptions near 1.0 mm with the location of the band minima being a sensitive marker of pyroxene geochemistry (Adams, 1974). The five filters between 860 nm and 1000 nm samples the 1.0 mm band at much higher resolution than any other region in the IMPÕs response range. This will allow a relatively sensitive determination of pyroxene mineralogy which can provide insight into geochemical history of the Martian crust.

Later in the mission much of the landing sitewill be imaged in all spectral channels. By that time we will have had the benefit of some chemical "ground truth" data from the APXS. Linking the chemical makeup of rocks and soil as determined by the APXS with the mineralogical interpretations of the reflectance spectra will be a powerful tool for classifying the local minerals.

The Pathfinder landing site was chosen to provide a variety of materials transported by floodwaters from as far away as the Martian highlands. Prior to the landing several observers including members of the IMP team will use the HST to produce multi-spectral maps of the landing area with high spatial resolution (20 km/pixel). The smallest IMP pixel will be approximately 1 mm allowing a view of the heterogeneity of the Martian surface to scales 8 orders of magnitude smaller than Earthbased observations. Spectral characteristics identified in the telescopic observations will be compared with the average "spectra" seen from the landing site (a 7 km diameter circle) providing regional context for the local spectra. Other supporting observations will be provided by the MGS orbiter when it starts returning data several months after the Pathfinder landing. Key instruments are the TES and MOC.

IMP Calibration Target Description

Calibration targets for the Imager for Mars Pathfinder provide a direct reference standard. There are two types of calibration targets: Color Targets, which provide color balance for true color images and a direct comparison of object and target for three iron oxides; and Radiometric Targets (RT), which permit radiometric calibration of scene elements. The calibration targets were made by direct casting of various pigments in a silicone binder; they were cast in aluminum molds with a bead-blasted surface that is nearly Lambertian. Targets were then "pre-yellowed" by ultraviolet exposure until the spectral behavior stabilized. Bi-directional reflectance measurements were performed using a Cary-14 spectrophotometer, retrofitted with improved hardware, electronics, and detector. Targets were measured from 320 nm to 1200 nm at 3.2 nm/channel, with incidence angle fixed at 0°, and emission angle fixed at 30°.

Two sets of color targets are on the Rover, and 4 sets are on the lander. The pigments for the 5 colors comprise three pure iron oxides: hematite (also provides red color reference), maghemite, and goethite and two additional color reference targets: Green (chromium dioxide paint pigment) and Blue (cobalt blue paint pigment). Reflectance spectra of the color targets are given in Figure 12.

There are 2 RTs on the lander, each consists of a white, gray, and black ring, with a shadow post at the center. Pigments used for the radiometric targets were: Rutile (white), carbon black (black), and a mixture of rutile and carbon black (gray). Reflectance spectra of the radiometric target A are shown in Figure 13, target B was created from the same mix and is essentially identical to A.

The primary requirement for multi-spectral interpretation of IMP data is well-calibrated image data. The calibration target set provides both a relative calibration reference as well as a standard for absolute reflectance. Procedures for using the calibration targets have been developed and tested in the Mars Garden, a simulated Martian landscape in close proximity to the calibration laboratory at the UA. The reflectance of the scene is calibrated by an examination of the RT nearby in time to the scene image. The counts for diffuse sky illumination will be measured by use of the shadows from the RT shadow post. Although care has been taken to reduce reflections and shadows from the spacecraft, these unwanted sources of light must be watched for. The corrected counts in the shadow region will be subtracted from the direct Sun + sky counts for each ring and the results will be plotted against the known reflectances of the target rings at the given filter bandpass. A linear, least-squares fit to the data will be performed which calculates a slope given in DN/sec/Reflectance Unit. The intercept term will be forced to zero and discrepancies provide a measure of the calibration uncertainty. If both the scene and target images have been corrected properly and scaled for differences in exposure time, incidence angle, and deviation from Lambert surface, the slope calculated for the target can be applied to the scene image giving absolute reflectance relative to an ideal Lambert surface.

Combining image data taken in different filters to produce reflectance spectra requires that images be correctly registered since pixel-to-pixel variations in reflectance are possible. Radiometric calibration is then performed based on the RT slope for that particular filter under the relevant environmental conditions. An example is shown in Figure 14 for an oxidized basaltic rock (ferric iron) imaged in the Mars Garden, and for a less oxidized basalt (ferrous iron). Clearly visible is the reddening of the rock as it oxidizes and the onset of the 860 nm ferric absorption band.

Studies of Aerosols and Water Vapor

The dust in the Martian atmosphere has been shown to play a crucial role in many atmospheric processes (Pollack et al., 1979) affecting the absorption of solar energy and the emission of thermal infrared radiation. Hence, the thermal structure of the atmosphere depends on the dust loading, the size of the particles, and their optical properties. The radiative heating and cooling both depends on the dust, and alters the forcing for atmospheric dynamics. Effects varying from local dust devils of short duration to dust storms that are essentially global and last for weeks can be observed. The dust in the atmosphere provides nuclei for condensation of water ice and even carbon dioxide ice at various altitudes and locations. In early morning hours before sunrise, ground fogs of water ice can form at low altitudes. In the polar regions, carbon dioxide ice can nucleate and grow to large sizes. Their precipitation out of the atmosphere is an important mechanism for removing dust from the atmosphere and depositing it, along with the ices, in the polar caps. In view of the large roles played by the dust and ice particles in the ther-mal balance, dynamics, and water and carbon dioxide transport, it is important to have an accurate picture of the distribution and properties of the dust and ice crystals at various times of day and year at many locations over the planet.

At a given location, the dust distribution can be characterized by number density as a function of altitude, the mean effective radius of the dust particles, their real and imaginary refractive indices as functions of wavelength, and a few additional param-eters that govern the shape of the single scattering phase function of the dust particles. For dust particles that are expected to be nonspherical, these ad-ditional parameters have been described by Pollack and Cuzzi (1980).

Thus, our science goals are to determine (1) the aerosol optical depth (as a function of wavelength) above the landing site, (2) the effective size and number density of the atmospheric aerosols (3) several parameters characterizing the shape of the aerosol particles (4) the vertical distribution of the aerosols; and (5) the imaginary refractive index of the aerosol particles. If both dust and ice particles are present, we hope to distinguish between them on the basis of their absorption properties as a function of wavelength. Finally, we plan to measure the variations in the aerosol properties with time by making regular observations throughout the mission.

Reduction and Analysis of Aerosol Observations

The aerosol optical depth will be calculated from the ratios of the flux from the sun obtained through different air masses. A sub frame (30 x 30 pixels) centered on the solar disk is returned for each observation, the background due to the surrounding sky brightness is subtracted, and the pixels covering the source are summed. When multiple scattering is negligible, the variation of flux from the external source obeys Beer's law, and the log of the ratio of intensities observed at two solar zenith angles divided by the difference of their secants gives the vertical optical depth of the aerosols. Departures from Beer's law at differ-ent times of day (after correction for the small ef-fects of multiple scattering) indicate changes in the vertical optical depth of the aerosols at the wavelength of the measurement. Color observations should be able to distinquish between the clouds and the dust. Viking imag-ing determined dust optical depths from a few tenths to values greater than 5 during global dust storms; however, their observations were limited to angles near the horizon, and they could observe the sun only in the red. We will have observations at four widely separated wavelengths (450, 670, 883, and 989 nm) using our solar filters (see Table 3); the mineralogical filters give complete wavelength coverage when we observe Phobos during the night.

The width of the forward scattering portion of the single-scattering phase function is controlled by diffraction and is determined by the projected area of the parti-cles with little sensitivity to particle shape (see Figure 15). Measurements of the variation of the sky intensity with azimuth when the sun was low in the sky were used by Viking to constrain the mean effective dust particle radius (Pollack et al., 1995). We will observe the solar aureole not only near sunrise and sunset, but at various other times of the day. The data will be collected at many wavelengths between 0.4 and 1.0 µm to provide an important consistency check of the particle size. The ability to point to high elevation angles permits aureole observations at night using Phobos as the source to directly measure the in-crease in size due to condensation of ice on dust nuclei during formation of early morning fogs.

Once the extinction optical depth as a function of wavelength is known, along with information on the shape of the phase function from the solar aureole (including side- and back-scattering observations) the size and cross section of the aerosol particles can be estimated. Division of the optical depth by the average cross-section gives the particle number density. Note that the optical depth as a function of wavelength provides an important consistency check on the particle size and shape, as the observations at each optical depth must give a consistent number density from the extinction data at all wavelengths.

The shape of the isophotes at greater distances from the sun constrains the middle and backscatter-ing por-tions of the phase function of the particles. This type of Viking imaging data has been analyzed to determine that dust above the Viking lander sites was domi-nated by particles having parameters more like flat plates than spheres, a fact taken to suggest the presence of clay-like ma-terials (Pollack et al., 1979). The same type of information will be available from our observa-tions at the Pathfinder landing site.

The brightness of the sky above the horizon after sunset (or before sunrise) is due to scattering by particles above a given altitude. The longer after sunset (or before sunrise), the higher in the atmosphere is the region probed. Once the size of the particles (and their forward scattering phase function) is known, the vertical distribution of the aerosols can be obtained from such observations. The dust above the Viking lander sites was relatively uniformly mixed with the gas up to altitudes of 50 km. Similar con-straints on vertical distribution will be available from our observations at the Pathfinder landing site.

Measurements of the Abundance of Water Vapor

The amount of water vapor in the atmosphere of Mars is known to vary significantly with time of day, location, and season (Jakosky and Farmer, 1982). Understanding the transport of water in the Martian atmosphere is a prerequisite for understanding the current and past climate of Mars and the evolution to the planet to its current state.

The brightness of the sun in a 5 nm wide water band will be measured after sunrise and before sunset when long atmospheric paths are available to constrain the atmospheric water content. Figure 16 indicates the transmission of the atmosphere as a function of water content at 10 airmasses (84° solar zenith angle) using line-by-line calculations based on the AFGL line list. The figure includes variations of a factor of 3 around the nominal wa-ter content of 10 pr µm. Absorptions of several percent will occur un-der these conditions. For 10 airmasses, measurements with good signal-to-noise (3 1000) provide good determinations of water vapor for modest dust abundances. Airmasses up to 25 would be available if the sun were visible at the local horizon and the dust load-ing were exceptionally small, yielding even greater sensi-tivity. For sufficiently thick dust layers, multiple scattering will complicate the analysis slightly. For dust optical depths up to unity below 10 airmasses, we have verified that the scattering effects are limited to only small angle scatter-ing, and that the water band/continuum ratio is equal to that expected for the case without dust to high accuracy.

It is important that the continuum filter be sufficiently close in wavelength to the water band filter so the variation in optical properties of the dust between the two filters is small. In Table 3, three continuum filters are listed (883, 925, and 989 nm) and a quadratic fit to the three intensity values allows an accurate measure of the continuum at the wavelengths of the water bands. Three water band filters are provided: two narrow (935 nm, left and right eyes) and one broadband (945 nm).

The measurements consist of image pairs of the sun in the three 5 nm wide filters obtained just after sunrise and before sunset each day. The relative response of the instrument in the filters will be well calibrated by images of the RT on the Lander. In addition, special care has been taken to control the temperature of the detector (heated to T=-20°C) during these observations to reduce variations due to changes in the relative response near the long wavelength end of the silicon CCD response. With stable temperatures and the target calibration, system-atic effects can be kept to low levels, and random noise is expected to dominate the accuracy of the water vapor determination.

Winds

The importance of wind interactions with the surface of Mars has been documented by telescopic studies of regional and global dust storms [e.g., Martin and Zurek, 1993; Zurek and Martin, 1993], by analysis of dune forms and transient albedo features seen from orbit [e.g., Sagan et al., 1972; Breed et al., 1979, Tsoar et al., 1979; Greeley and Iversen, 1985; Greeley et al., 1992; Thomas and Gierasch, 1995], and by observation of wind-shaped drifts at the Viking Lander sites [e.g., Sharp and Malin, 1984]. Both the occurrence of wind and the presence of fine particles subject to aeolian transport are reasonable expectations for the Pathfinder landing site. Despite the randomness inherent in turbulent atmospheric motion, with time-averaging the behavior of wind close to the surface is reasonably ordered.

Surface irregularities drag on moving air, slowing its passage over the surface, and this momentum absorption propagates upward through the atmosphere to produce a wind profile that is concave upward. In the absence of thermal influences the time-averaged wind profile can be described logarithmically (Prandtl, 1925)

u = (u*/k) ln(z/z₀) (1)

where u = wind speed at height z, k = von Karman's constant (usually assigned a value of 0.4), z₀ = aerodynamic roughness, and u* = friction speed. The friction speed = (t/r)1/2, where t = shear stress on the surface due to wind, and r = atmospheric density. Aerodynamic roughness, z₀ , is an intrinsic property of the surface and is invariant with wind speed (although not necessarily with wind direction). Wind friction speed, u*, is an indication of the drag on the surface due to wind and is the most important indicator of the potential of wind for moving fine particles; most expressions derived for wind-driven particle flux are functions of u*3 [Bagnold, 1941]. Together, u* and z₀ describe the wind profile at any time and its potential to move particles across the surface. However, measurements of wind speed at more than one height are required for u* and z₀ to be determined unambiguously; this was not possible with Viking Lander data.

Three windsocks have been developed for attachment to the one-meter mast of the Atmosphere Structure Instrument/Meteorology (ASI/Met) package. The windsocks will be imaged repeatedly by IMP (with sub-framing and compression) over short time intervals to allow measurement of the wind speed boundary-layer profile, including determination of aerodynamic roughness, wind friction speed, and shear stress on the surface due to wind. The ability to determine these parameters allows evaluation of the potential of wind to move fine particles across the landing site. Changes in the surface distribution of mobile materials during the mission can be related to z₀ and u* determined from IMP windsock images. Surface changes and winds at the landing site might also be related to wind streaks and other albedo features within the landing ellipse seen in Viking Orbiter images.

Each windsock consists of an aluminum cone rigidly joined to a steel and aluminum counterweight spike which together pivot on a small, low-friction gimbal mount (Figure 17). Each windsock assembly is less than 10 cm long and is mounted (flexibly) at the end of a support strut extending 10 cm from the ASI/Met mast. The windsocks are counter-balanced for sensitivity to wind at typical Martian surface pressures. The design was tested at one atmosphere in the wind tunnel and in the field, and at equivalent Martian atmospheric pressure in the ASU low-pressure wind tunnel at NASA-Ames, and found to be aerodynamically stable at all deflection angles. All windsock materials are fully conducting and grounded to prevent static charge affecting windsock deflection. Each windsock assembly can withstand structural accelerations of several hundred Gs; the three windsocks are mounted at heights of 33.1, 62.4, and 91.6 cm.

Three flight models and three flight spares were calibrated at one atmosphere and at equivalent Martian atmosphere. At one atmosphere each flight unit was imaged simultaneously from horizontal and vertical directions at seventeen deflection angles at seven strut azimuths, and at seven deflection angles at five additional strut azimuths. At Martian equivalent atmospheric pressure, each flight unit was similarly imaged at an average of eleven deflection angles at seven strut azimuths. Some minor hysterisis (1-2 degrees in deflection) was encountered during early experiments, so all data were obtained at both increasing and decreasing wind speeds. Results for windsock flight unit 1 at one atmosphere and at equivalent Martian atmosphere for the representative strut azimuth of 60° are summarized in Figure 18, deflections are measured from vertical. Dead zone deflections (neutral, no wind) of 5-6 degrees indicated in Figure 18 are representative of other azimuths.

Wind speeds and deflections at one combination of atmospheric density and surface gravity can be related to wind speeds and deflections at other conditions by equating torque from aerodynamic forces causing deflection with torque from gravitational forces resisting deflection:

(2) where u = wind speed, R1 = distance between pivot and center of mass, M = non-counter-balanced mass, g = acceleration of gravity, q = deflection from vertical, R2 = distance between pivot and center of aerodynamic pressure, Ad = effective aerodynamic cross-section, and r = atmospheric density (a function of pressure and temperature). As an example, Equation (2) was applied to data obtained at one atmosphere (bottom curve in Figure 18 to predict performance under Martian conditions of 8 mb, 225 K, and 3.7 m/sec2 (top curve in Figure 18). Predictions for the same Martian conditions were calculated by applying Equation (2) to data obtained using air at ~15 mb. Overlap of the two predictions in Figure 18 indicates that expected sensitivities for these Martian conditions will be from about 5-30 m/sec.

Magnetic Properties Experiment

A remarkable result from the Viking missions was the discovery that the Martian soil is highly magnetic, in the sense that the soil is attracted by small permanent magnets (Hargraves et al., 1977, 1979). The result was remarkable in that similar looking red soils on Earth are seldom as magnetic as the soil found on Mars. At both landing sites the Martian soil adhered almost equally well to a strong and a weak Sm--Co magnet attached to the Viking lander backhoe. The strong magnet had a surface magnetic field and a surface magnetic field gradient of 0.25 T and 100 T/m respectively. The corresponding numbers for the weak magnet were 0.07 T and 30 T/m. Based on the pictures of the amount of soil clinging to the magnets, the particles in the Martian dust contain between 1% and 7% of a strongly magnetic phase and limits for the saturation magnetization ss were implied:

Chemical analyses by means of the Viking X-ray fluorescence spectrometer indicated a content of about 18% Fe2O3 by weight. The saturation magnetization may then be bounded:

Therefore, since both magnets were saturated with magnetic soil throughout the Viking mission, Mars Pathfinder carries three permanent magnets of lesser strengths.

Besides the backhoe magnets (strong and weak) each Viking lander carried a so called Reference Test Chart (RTC) magnet, which was exposed exclusively to the airborne dust. The RTC magnets were of the strong type only and attracted a substantial amount of airborne dust. The dust on the RTC magnets, the dust on the backhoe magnets and the dust on the surface on Mars were optically identical.

During the favorable opposition in 1988, the reflection spectra of bright regions of Mars showed that the Fe(III) in the Martian fines is found partly in so called nanocrystals (linear dimensions below about 10 nm), and partly in macroscopic crystals. Most of the Fe(III) ions seem to occur in nanometer-sized particles (Morris et al., 1989, Morris and Lauer, 1990, Bell et al., 1990, Morris et al., 1990, Soderblom, 1992). In large nanophase particles (>20 nm) the magnetization vector M is locked into a so-called easy direction of magnetization by the forces of anisotropy. But for smaller nanophase particles above a certain temperature the anisotropy forces cannot hold the magnetization vector M fixed in an easy direction, and M fluctuates stochastically in response to the thermal noise. The phenomenon is called superparamagnetism (O'Reilly, 1984). Besides, antiferromagnetic particles (for instance, hematite) develop a substantial spontaneous magnetization when the particle size is small enough (d < 10 nm).

The dominant iron species in rock-forming silicate minerals is Fe(II). From optical spectroscopy it is known that the reddish color of the Martian soil is caused by Fe(III) compounds; therefore, an oxidation of the surface rocks of Mars has taken place. The critical question is whether the reddish soil of Mars formed billions of years ago when the planet was warmer and wetter, or whether the soil-forming process on Mars still going on today through the direct interaction of an oxidizing atmosphere with the surface rocks. The break up of H2O by ultraviolet light provides a source of highly reactive radicals that may combine with -- and oxidize -- surface materials, among these the Fe(II) compounds. Oxides of Fe(III) formed via direct solid--gas reactions will generally be different from iron oxides and iron hydroxides formed via precipitation in abundant liquid water. Two major pathways exist for the weathering process on Mars, and thus for the formation of the magnetic phase in the weathering products.

Precipitation in Liquid Water

The first major pathway involves the action of abundant liquid water. The ions -- including Fe(II) -- of the primary rocks may be completely released into aqueous solution. The soluble Fe(II) is then oxidized to Fe(III), which is insoluble in water with pH > 3. The parameters that determine the nature of the precipitate are numerous and the highly complex processes are not understood in detail even for precipitation processes studied under laboratory conditions (Banin et al., 1993). Here we limit ourselves to a brief summary of some results relevant for the study of the magnetism in the soil.

1. The precipitate formed from iron dissolved in water will be pure iron hydroxides and iron oxides.

2. If oxygen is comparatively scarce the oxidation of Fe(II) will be slow, and magnetite (Fe3O4) or maghemite (g-Fe2O3) will often be the end products (strongly magnetic).

3. If oxygen is abundant with a swift oxidation of Fe(II), hematite (a-Fe2O3) and goethite (a-FeOOH) will form (less magnetic).

Most often the red tropical soils on Earth contain a substantial amount of superparamagnetic (nanophase) particles. The conditions under which superparamagnetic particles form in nature are not known; however, a magnetic phase formed via precipitation in liquid water will not contain titanium.

Gas­Solid Reactions

In the second major pathway for the formation of the magnetic phase the weathering is assumed to be caused by the direct interaction of an oxidizing atmosphere with the solid rocks. The SNC meteorites are believed to be surface rocks from Mars (McSween, 1985) and they appear to support this interaction. The Shergottites and Nakhlites are basaltic rocks containing an abundance of about 2% of a ferromagnetic mineral, titanomagnetite (Fe3-xTixO4, 0.1 < x < 0.6; Dodd, 1981).

The present surface conditions on Mars are drastically different from those on Earth. The soil forming transformations on Mars must nevertheless be proceeding in basically the same sense. Basaltic rocks are tending to equilibrium in conditions of lower temperature and pressure and significantly higher oxygen fugacity than those prevailing when the basalts originally formed. Therefore, if weathering on Mars has been less intense, i.e. if the Fe(II) in the bedrock was never completely released into aqueous solution, the primary titanomagnetite grains in the surface rocks of Mars may have survived intact, perhaps becoming oxidized to titanomaghemite. Such residual magnetic grains­inherited from the underlying rocks­will be attracted to the permanent magnets, and in this case the magnetic grains will evidently contain the element titanium. From X-ray fluorescence performed during the Viking missions the element titanium is known to be present in the Martian soil in an abundance of about 0.9% by weight (Toulmin et al., 1977).

Magnetic Minerals

The magnetism of strongly magnetic terrestrial soils (ss > 1 Am2/kg) is caused by the presence of one or more of the magnetic oxides. Table 5 gives a list of well known minerals that carry a spontaneous magnetization below 295K. The magnetism of soils is more complicated than one would expect from the listing of magnetic minerals in Table 5. Isomorphous substitutions (Ti, Al) alter the magnetic properties of the compounds drastically. Another complication is introduced by small particle size because of the onset of superparamagnetism. From the table it will be seen that nanophase (superparamagnetic) particles of hematite may be an order of magnitude more magnetic than macroscopic particles of hematite. Nanophase hematite has been suggested as the dominant material in the soil of Mars (Morris et al., 1989). A further complication for the study of the magnetism on Mars is the fact that the magnetic phase(s) in the soil may occur in composite (multi--phase) particles, for instance, as a pigment dispersed throughout the volume of silicate particles (e.g. smectite clay), or as a coating on silicate particles.

In summary, the main scientific goals for the Magnetic Properties Experiments on Mars Pathfinder are (1) To determine the magnetization of the Martian dust with higher precision than was done during the Viking mission (2) To identify the mineral(s) responsible for the magnetization of the Martian soil and to determine their abundance (3) To establish if the magnetic phase(s) occur as discrete particles or as part of composite particles and (4) To obtain information about the size of the particles attracted by the magnets.

The Magnetic Instruments and their Operations

The Mars Pathfinder Lander carries the following types of instruments: two Magnet Arrays, one Tip Plate Magnet, and two Ramp Magnets which are each described briefly.

Magnet Array

A Magnet Array (MA) consists of two blocks of magnesium with permanent Sm-Co magnets embedded. One block of the MA carries two magnets of Viking strength and the other block carries three magnets of lesser strength. For protection against corrosion of the magnesium a thin layer (~500 nm) of platinum has been sputtered onto the surface. This produces an electrically conductive surface, which is gray and gives a good visible contrast to iron oxides. Figure 19 shows a sketch of an MA with typical values of the magnitude of the magnetic field gradient ÑB and magnetic field B at the surface of the instrument. The magnets are constructed to give a well defined magnetic field gradient between an outer ring and an inner cylinder, which have opposite polarity.

As an example, Figure 20 shows the variation of the magnetic field B across the strongest magnet of a Magnet Array. The field has cylindrical symmetry and the figure shows the component Bz, perpendicular to the surface and the component Br in the plane of the surface of the instrument. The radius r is measured from the symmetry axis of the cylindrical magnet. Figure 21 shows the corresponding values for the magnetic field gradient. The force on a given particle is proportional to the magnetic field gradient.

The MAs are mounted on the lander at different heights above the ground surface both passively exposed to wind blown particles. The lower MA is mounted on the central petal so low that it may attract saltating sand grains. The MAs will be periodically viewed by the IMP using subframing to reduce the data volume. Table 6 lists distances from the IMP to the Magnet Arrays and the corresponding resolutions obtained for both deployed and stowed positions of the IMP. In some cases, parts of the MAs will be obscured by the high gain antenna and the electronics box on the lander.

An important scientific goal of the Magnet Array is to determine the saturation magnetization of the Martian dust with a precision higher than was obtained during the Viking mission. Figure 22 shows results of calculations on how the dust will adhere to the various magnets. Besides the magnetization, the size of the particles is an important parameter. The field gradient -- and hence the force -- decreases strongly with z, the height above the surface of the instrument so that the average force on a large particle is smaller than the average force on a small particle with the same magnetization. The wind velocity will also influence the dust covering of the magnet; the calculations illustrated in Figure 22 were performed assuming a particle size of 1 µm and a wind velocity of 3 m/sec. Figure 23 shows the results of an experiment performed with selected Mars Sample Analogues with material blown directly onto the surface of the magnets. The strongly magnetic mineral maghemite sticks to all five magnets. Hematite, a canted antiferromagnet, sticks to the strongest magnet only (compare to figure 22). This illustrates the fact that hematite particles with linear dimensions of more than 10 nm cannot be solely responsible for the magnetization of the Martian soil. Both the weak and the strong Viking backhoe magnets were saturated with magnetic dust. The effect of particle size can be seen by comparing the patterns from black sand from Anholt with those from maghemite, which has a nearly identical value of the spontaneous magnetization. The magnetic particles in the Anholt sand are larger (~500 µm) than the particles in the maghemite sample (~1 µm) and therefore only very few grains of the black sand stick to the weakest magnet.

Tip Plate Magnet

The TPM consists of one magnet assembly embedded in magnesium. This instrument is constructed in such a way that the magnitude of the magnetic field and the field gradient vary across the surface that is exposed to airborne dust; see Figure 24. The highest values of the magnetic field gradient and the magnetic field strength are 130 Tm-1 and 0.35 T (stronger than the strongest Viking magnet). The tip plate rigidly connects the upper ends of the 3 longerons to the base flange of the camera head and will be 1.6 m above the surface when the IMP is in the deployed position. The TPM is placed about ~10 cm from the eye of the IMP and the magnetic dust collected from the air will be imaged through a diopter lens. This allows a resolution of 0.18 mm/pixel. Tests have shown that the depth of field of the IMP is so large that imaging of the TPM using the spectroscopic filters is possible with a reduction in resolution. The high resolution in the images of the dust on the Tip Plate Magnet makes it possible to deduce if the magnetic grains align in visible chains. The grains may form chains when they are essentially single phase magnetic particles with a substantial magnetization. Multi-phase particles­or superparamagnetic particles­will generally not form chains. The exact conditions for spontaneous formation of visible chains is still under investigation. The TPM experiment will address all the scientific goals 1 through 4. Figure 25 shows two selected Mars Sample Analogues on the TPM. The spacing between the chains depends on the magnetization of the individual grains, which in turn is a function of both their magnetic properties and of the magnetic field the chains are subjected on the TPM. A correlation of the distribution of chain structures with the local values of the magnitude of the magnetic field gradient and the magnetic field will give information on the value of s of the individual grains and on the form of their magnetization curve. The Ramp Magnet On the Rover petal of the Mars Pathfinder lander there will be two deployable ramps one of which the Rover will descend to the surface. At the outer end of each ramp is fastened a so-called Ramp Magnet of size 10 x 10 cm. The Ramp Magnets are constructed from strips of magnetic material that are covered by aluminized Mylar. This construction allows the Ramp Magnets to be rolled up into a cylinder with a diameter of 2.5 cm to fit into the undeployed ramps. When deployed, each Ramp Magnet will form a plane surface with average values of the magnitude of the magnetic field gradient and the magnetic field corresponding roughly to values between those of magnet number 3 and magnet number 4 of an MA. The deployment of the ramps and the motions of the Rover will raise magnetic dust, some of this dust will be caught by the Ramp Magnets that are constructed to perform a magnetic fractionation of the dust. The Ramp Magnets will not attract composite (multi-phase) particles very efficiently. If, however, the magnetism of the soil is caused by discrete (single-phase) particles of a strongly magnetic oxide, the Ramp Magnet will attract the magnetic part of the dust. This is true even if all the Ti (0.9% of soil by weight) in the soil is found in the magnetic phase. Titanomagnetite or titanomaghemite, with about 20% Ti will be sufficiently magnetic (~50 Am2/kg) to be attracted by the Ramp Magnet. The Ramp Magnets will be periodically viewed by the IMP to ascertain if dust is present. If so, the Rover will return to at least one of the ramps and perform an APXS analysis of the dust on the Ramp Magnet. In this way the question of the possible enrichment of the element Ti in the magnetic phase will be addressed. ACKNOWLEDGMENTS Developing a camera concept into a spaceflight-ready instrument requires the concerted efforts of a large number of dedicated people and the support of many organizations. We have the pleasure to have worked with many talented individuals and are deeply grateful for their fine efforts. Local engineering support at the UA is provided by Robert Reynolds and Roger Tanner in mechanical, thermal, and calibration; Steve Bell in electronics; Chris Shinohara, Terry Friedman, and Lyn Doose in software; and Jon Weinberg, Greg Hoppa, Nancy Chabot, Robert Marcialis, Mike Burkland, Robert Singer, and James Palmer in calibration and data reduction. In addition, the project could not have proceeded without the support of Mary Williams and Maribel Cota in the project office. The JPL project office, lead by Anthony Spear, provided the services of John Wellman and Tom Tomey to guide the integration and testing during all stages of the project; they and the rest of the Pathfinder team have been a pleasure to work with. Lockheed Martin Astronautics gave a large measure of engineering support both in the proposal preparation, lead by Joseph Martin, and in the detailed engineering of the camera system. Dave Barnett was the program manager and his engineering team consisted of John Montgomery, Mark Bigler, and Mike Pollard. The hardware development at MPAe was managed by Detlef Koschny and the electronic design and testing of the CCDs was achieved due to the tireless efforts of Rainer Kramm. Frank Rabe helped develop the data compression software in Braunschweig, Germany. Wes Ward at USGS and G. Wilson at ASU contibuted to the program. A special thanks goes to Ken Herkenhoff for the tremendous support he has given us as the JPL instrument representative. The magnetic properties experiment is supported by the Danish Natural Science Research council, the Ministry of Research and the Ib Henriksen Foundation. We are grateful to James Bell, Matt Golombek, and Thomas McCord for their thoughtful reviews.

Table 1.

REFERENCES Adams, J.B., Visible and near-IR diffuse reflectance spectra of pyroxenes as applied to remote sensing of solid objects in the solar system, J. Geophys. Res. 79, 4829-4836, 1974. Adams, J. B., and McCord, T. B. , Mars: Interpretation of spectral reflectivity of light and dark regions, J. Geophys. Res., 74, 4851-4856, 1969. Arvidson, R.E., Guinness, E.A., and Lee, S., Differential aeolian redistribution rates on Mars, Nature, 278, 533, 1979. Arvidson, R.E., Gooding, J.L., and Moore, H.J. The Martian surface as imaged, sampled, and analyzed by the Viking landers, Rev. Geophys.,27, 39, 1989 Bagnold, R. A., The Physics of Blown Sand and Desert Dunes, Chapman and Hall, London, 265 pp., 1941. Baird, A. K., and Clark, B. C. , On the original igneous source of Martian fines, Icarus, 45, 113-123, 1981. Banin, A., R. Ben-Shlomo, L. Margulies, D. F. Blake, R. L. Mancinelli, and Gehring, A. U. , The nanophase iron mineral(s) in Mars Soil, J. Geophys. Res., 98, 20,831-20,853, 1993. Bell, J. F., III, T. B. McCord, and Owensby, P. D. , Observational evidence of iron oxides on Mars, J. Geophys. Res., 95, 14,447-14,461, 1990. Bell, J. F., III, R. V. Morris, and J. B. Adams, Thermally altered palagonite tephra: A spectral and process analogue to the soil and dust of Mars, J. Geophys. Res., 98, 3373-3385, 1993. Binder, A.B., Arvidson, R.E., Guinness, E.A., Jones, K.L., Morris, E.C., Mutch, T.A., Pieri, D.C., and Sagan, C., The geology of the Viking Lander 1 site. J. Geophys. Res., 82, 4439-4451, 1977. Breed, C. S., Grolier, M. J., and McCauley, J. F. , Morphology and distribution of common "sand" dunes on Mars: Comparison with the Earth, J. Geophys. Res., 84, 8183-8204, 1979. Burns, R. G., Does feroxyhyte occur on the surface of Mars?, Nature, 285, 647, 1980. Christensen, P.R., and Moore, H.J., The martian surface layer, in Mars, Kieffer, H.H., Jakosky, B.M., Snyder, C.W., and Matthews, M.S., eds., Univ. of Arizona Press, Tucson, 686-729, 1992. Coey, J. M. D., Mørup, S., Madsen, M. B. , and Knudsen, J. M. , Titanomaghemite in magnetic soils on Earth and Mars, J. Geophys. Res., 95, 14,423-14,425, 1990. Dodd, R. T., Meteorites, A petrologic-chemical synthesis, Cambridge University Press, 1981. Giese, B., Oberst, G., Kirk, R., and Zeitler, W., The topography of asteroid Ida: A comparison between photogrammetric and two-dimensional photoclinometric image analysis, Proc. ISPRS XVIII, Vienna, Austria, 1996. Gooding, J. L., Soil mineralogy and chemistry on Mars: Possible clues from salts and clays in SNC meteorites, Icarus, 99, 28 41, 1992. Greeley, R. and Iversen, J. D. , Wind as a Geological Process, Cambridge, London, 1985. Greeley, R., Lancaster, N., Lee, S., and Thomas, P. , Martian aeolian processes, sediments, and features, in Mars, Kieffer, H. H., Jakosky, B. M., Snyder, C. W., and M. S. Matthews, eds., University of Arizona, Tucson, 730-766, 1992. Hargraves, R. B., Collinson, D. W. , Arvidson, R. E. , and Spitzer, C. R. , Viking magnetic properties investigation: Further results, Science, 194, 1303-1308, 1976. Hargraves, R. B., Collinson, D. W. , Arvidson, R. E. , and Spitzer, C. R., The Viking magnetic properties experiment: Primary mission results, J. Geophys. Res., 82, 4547-4558, 1977. Hargraves, R. B., Collinson, D. W. , Arvidson, R. E., and Gates, P. M. , Viking magnetic properties experiment: Extended mission results, J. Geophys. Res., 84, 8379-8384, 1979. Jakosky, B.M., and Farmer, C.B., The seasonal and global behavior of water vapor in the Mars atmosphere: Complete global results of the Viking atmospheric water detector experiment, J. Geophys. Res. 87, 2999-3019, 1982. Kirk, R.L., A fast finite-element algorithm for two-dimensional photoclinometry, Ph.D. dissertation, California Institute of Technology, Part III, 165-258, 1987. Martin, L. J., and Zurek, R. W. , An analysis of the history of dust activity on Mars, J. Geophys. Res., 98, 3221-3246, 1993. McSween, H. Y. SNC meteorites: Clues to Martian petrologic evolution, Rev. of Geophysics, 23, 391-416, 1985. Moore, H.J., and Jakosky, B.M., Viking landing sites, remote-sensing observations, and physical properties of Martian surface materials, Icarus, 81, 164, 1989. Morris, R.V., Lauer, H.V. Jr., Lawson, C.A., Gibson, E.K. Jr., Nace, G.A., and Stewart, U., Spectral and other physicochemical properties of submicron powders of hematite, maghemite, magnetite, goethite, and lepidocrocite, J. Geophys. Res. 90, 3126-3144, 1985. Morris, R. V., Agresti, D. G. , Lauer, Jr., H. V. , Newcomb, J. A. , Shelfer, T. S. , and Murali, A. V. , Evidence for pigmentary hematite on Mars based on optical, magnetic and Mossbauer studies of superparamagnetic (nanocrystal]ine) hematite, J. Geophys. Res., 94, 2760-2778, 1989. Morris R. V. and Lauer, H. V. , Matrix effects for reflectivity spectra of dispersed nanophase (superparamagnetic) hematite with application to Martian data, J. Geophys. Res., 95, 5101-5109, 1990. Morris, R. V., Gooding, J.L., Lauer, H. V. , and Singer, R. B. , Origins of the Mars-like spectral and magnetic properties of a Hawaiian palagonitic soil, J. Geophys. Res., 95, 14,427-14,434, 1990. Mutch, T.A., Arvidson, R.E., Binder, A.B., Guinness, E.A., and Morris, E.C., The geology of the Viking Lander 2 site, J. Geophys. Res., 82, 4452-4467, 1977. O'Reilly, Rock and mineral magnetism, Blackie, Glasgow and London, 1984. Pollack, J.B., D.S.Colburn, F.Michael, R.Kahn, C.E.Carlston, and D.Pidek, Properties and effects of dust particles suspended in the Martian atmosphere, J. Geophys. Res. 84, 2929-2945, 1979. Pollack, J.B. and J.N.Cuzzi, Scattering by nonspherical particles of size compara-ble to a wavelength: A new semi-empirical theory and its application to tropo-spheric aerosols, J.A.S, 37, 868-881, 1980. Pollack, J.G., D.Colburn, R. Kahn, J.Hunter, W.Van Camp, C.E.Carlston, and M.R.Wolf, Properties of aerosols in the martian atmosphere, as inferred from Viking lander imaging data, J. Geophys. Res. 82, 4479-4496, 1977. Pollack, J.B., Ockert-Bell, M.E., and Shepard, M.K., Viking Lander image analysis of Martian atmospheric dust, J. Geophys. Res. 100, 5235-5250, 1995. Prandtl, L., Uber die ausgebildete Turbulenz, ZAMM, 5, 136-139, 1925; and Proc. 2nd. Intern. Congr. Appl. Mech., Zurich, 62-75, 1926. Rueffer, P., Rabe, F., and Gliem, F., Enhancement of IMP lossy image data compression using LCT, Space and Earth Science Data Compression Workshop, JPL Publication 95-8, 1995. Sagan, C., Veverka, J., Fox, P., Dubisch, R., Lederberg, J., Levinthal, E., Quam, L., Tucker, R., Pollack, J. B., and B. A. Smith, Variable features on Mars: Preliminary Mariner 9 results, Icarus, 17, 346-372, 1972. Sagan, C., Pieri, D., and Fox, P., Particle motion on Mars inferred from the Viking Lander cameras, J. Geophys. Res. ,82, 4430-4438, 1977. Schlichting, H., Boundary Layer Theory, McGraw-Hill, New York, 7th ed., 578-595, 1979. Sharp, R. P., and M. C. Malin, Surface geology from Viking Landers on Mars: A second look, Geol. Soc. Amer. Bull., 96, 1398-1412, 1984. Sherman D. M., R. G. Burns, and V. M. Burns, Spectral characteristics of the iron oxides with application to the Martian bright region mineralogy, J. Geophys. Res., B87, 10,169-10,180, 1982. Sherman, D. M., and T. D. Waite, Electronic spectra of Fe3+ oxides and oxide hydroxides in the near IR to near UV, American Mineralogist, 70, 1262-1269, 1985. Singer, R. B., T. B. McCord, R. N. Clark, J. B. Adams, and R. L. Huguenin, Mars surface composition from reflection spectroscopy: A summary, J. Geophys. Res., 84, 8415-8426, 1979 . Singer, R. B., Spectral evidence for the mineralogy of high-albedo soils and dust on Mars, J. Geophys. Res., B87, 10,159-10,168, 1982. Soderblom, L. A., The composition and mineralogy of the Martian surface from spectroscopic observations: 0.3 mm to 50 mm. in Mars edited by H. H. Kieffer, B. M. Jakosky, C. W. Snyder and M. S. Matthews pp. 557-593. The University of Arizona Press, Tucson and London, 1992. Thomas, P. C. and Gierasch, P. J., Polar margin dunes and winds on Mars, J. Geophys. Res., 100, 5397-5406, 1995. Tomasko, M.G.,Doose, L.R.,Smith, P.H., Fellows, C.,Rizk, B., See, C., Bushroe, M., McFarlane, E., Wegryn, E., Frans, E., Clark, R., Prout, M., and Clapp, S., The Descent Imager/Spectral Radiometer (DISR) aboard the Huygens Probe of Titan, SPIE Vol. 2803, 64-74, 1996. Toulmin, P., Baird, A. K., Clark, B. C., Keil, K., Rose, H. J., Christian, R. P., Evans, P. H., and Kelliher, W. C., Geochemical and mineralogical interpretation of the Viking inorganic chemical results, J. Geophys. Res., 82, 4625-4634, 1977. Tsoar, H., Greeley, R., and Peterfreund, A. R., Mars: The north polar sand sea and related wind patterns, J. Geophys. Res., 84, 8167-8180, 1979. Vasavada, A.R., Milavec, T.J., and Paige, D.A., Microcraters on Mars: Evidence for past climatic variations, J. Geophys. Res. 98, 3469-3476, 1993. Wolff, L.B., and Angelopolou, E., Three-dimensional stereo using photometric ratios, in Computer Vision-ECCV Ô94, Third European Conference on Computer Vision, Stockholm, Sweden, Proc. V. II, 247-258, May 1994. Zurek, R. W., and Martin, L. J., Interannual variability of planet-encircling dust storms on Mars, J. Geophys. Res., 98, 3247-3259, 1993.