Performance validation of the ExoMars 2018 WISDOM GPR in ice caves, Austria

Author’s Accepted Manuscript Performance validation of the ExoMars 2018 WISDOM GPR in ice caves, Austria S. Dorizon, V. Ciarletti, D. Plettemeier, W.S. Benedix www.elsevier.com

PII: DOI: Reference:

S0032-0633(15)00307-4 http://dx.doi.org/10.1016/j.pss.2015.10.008 PSS4080

To appear in: Planetary and Space Science Received date: 28 January 2015 Revised date: 16 October 2015 Accepted date: 20 October 2015 Cite this article as: S. Dorizon, V. Ciarletti, D. Plettemeier and W.S. Benedix, Performance validation of the ExoMars 2018 WISDOM GPR in ice caves, A u s t r i a , Planetary and Space Science, http://dx.doi.org/10.1016/j.pss.2015.10.008 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Performance validation of the ExoMars 2018 WISDOM GPR in ice caves, Austria S. Dorizona,b , V. Ciarlettia,b , D. Plettemeierc , W.S. Benedixc a.

Université Versailles Saint Quentin en Yvelines, 45 avenue des Etats Unis - 78035 Versailles cedex, France

b.

Laboratoire atmosphères, Milieux, Observations Spatiales (LATMOS), Quartier des Garennes, 11 bd d’Alembert - 78280 Guyancourt, France

c.

Technische Universität Dresden, TU Dresden, Institut für Nachrichtentechnik, Lehrstuhl Hochfrequenztechnik - 01062 Dresden, Germany

Corresponding author: Sophie Dorizon, LATMOS-OVSQ, Quartier des Garennes, 11 bd d’Alembert - 78280 Guyancourt, France E-mail : [email protected] Phone : +33 (0)180285092

Abstract The WISDOM (Water Ice Subsurface Deposits Observations on Mars) Ground Penetrating Radar has been selected to be part of the ExoMars 2018 exobiological rover mission. A prototype has been tested during the Mars Simulation organized by the Austrian Space Forum in Alpine ice caves in Dachstein, Austria. This campaign provided the opportunity to validate methods developed to process WISDOM’s data in a well-documented environment and to retrieve geometrical and quantitative information about the 3D structure and the electromagnetic properties of the subsurface. We estimate the ice thickness in different locations inside the ice caves, and show that this ice is formed of fine strata with different properties. Data analysis allows reconstructing the bedrock in a 3D environment where a complete survey was performed.

Keywords ExoMars, WISDOM, GPR, subsurface, ice caves, permittivity 1

1. Introduction The WISDOM (Water Ice Subsurface Deposits Observations on Mars) Ground Penetrating Radar (GPR) has been selected to be part of the European Space Agency (ESA) ExoMars 2018 exobiological rover mission (Barnes et al., 2006; Bousquet et al., 2012). The main scientific objectives of the mission are to (i) search for traces of past or present life, (ii) characterize the geologic and geochemical environment and distribution of water as a function of depth in the shallow subsurface and (iii) investigate Mars’ subsurface and understand its evolution and habitability.

Figure 1 : ESA's ExoMars Rover. Credit: ESA The WISDOM GPR (Ciarletti et al., 2011) will be accommodated aboard the ExoMars rover (Fig.1) and will be used to perform large-scale scientific investigations of the shallow subsurface. This instrument is designed to explore the first ~3 meters of the subsurface with a vertical resolution of a few centimeters. It will detect and characterize buried structures such as sedimentary layers, rocks, potential massive ice deposits and transient or persistent occurrences of liquid water, and consequently provide high resolution description of the structure, stratigraphy and nature of the shallow subsurface. WISDOM will also be a strong help to guide and secure the drilling operations (~up to 2 meters depth) to locations of interest 2

by determining the nature, depth and extent of potential targets and by identifying potential hazards. WISDOM data will thus provide the first hand information about the Martian shallow subsurface, a domain almost unexplored until now. The results will be crucial to understand the processes and environmental conditions that lead to the actual geological environment, as well as the past and present habitability of the planet. In spring 2012, the WISDOM instrument team participated in the Dachstein Mars Simulation 2012 organized by the Austrian Space Forum (Groemer et al., 2012) in the Alpine ice caves in Dachstein, Austria. Two identical prototypes of the WISDOM GPR were operated on different platforms. Radar investigations were performed in four different areas into the caves. While this icy environment might not be the best Martian analog, it still is a reasonably liquid water-free radar target and, most of all, it provides an excellent opportunity to validate methods developed to process WISDOM’s data. Indeed, GPR can be successfully used to map the thickness and structure of the ice, as demonstrated by previous studies (Gogineni et al., 2001, Plewes and Hubbard, 2001, Arcone et al., 1998). We first briefly describe the geological context and the experimental set-up for the field test. After an overview of the WISDOM GPR capacities and functionalities, two independent methods are presented and applied on the experimental data in order to retrieve quantitative information about the 3D structure of the subsurface and its electromagnetic properties. Eventually, we discuss our results in light of previous studies performed in the same cave system (Behm and Haussman, 2008, Haussman and Behm, 2011).

3

2. The Dachstein field test The Dachstein massif is situated in the Northern Calcareous Alps and covers an area of around 20×30 km, permeated by a rich cave system, including some of the largest caves in Austria. Caves are widely covered with ice with a thickness ranging from few centimeters up to several meters, either starting at the surface or buried under limestone. The investigated area during the 2012 Mars simulation is the Giant Ice Cave (Rieseneishöhle) with a passage length of 2.7 km. The elevation of this location is 1460 m above sea level, and observations in Tristandom, which is the first cavity when entering into the Giant Ice Cave, shows that the maximum ice thickness is about 10 m (Groemer et al., 2012). However, the diversity inside this cave is important and as passing the different rooms, ice thickness appears to vary from few dozens of centimeters to five meters. As part of the AUSTRO ICE CAVES 2100 project, Haussmann and Behm’s study in 2006 partially took place in this Giant Ice Cave. Even if acquisitions were not recorded at the exact same place as WISDOM (Behm and Haussmann, 2008, Haussmann and Behm, 2011), we can assume, since the ice formation processes should have been quite similar despite the local differences (mostly linked to the elevation), that our results will have similitudes with their conclusions. Based on these information and direct observations, we thus anticipated the ice sheet cover to be in the range of a few dozens of centimeters to a few meters with an internal layering composed of few-centimeters thickness strata, sitting on top of a fractured bedrock formed of boulders which dimensions can vary from a few centimeters to one or two meters. During the Mars simulation field test, over ~2400 single acquisitions were performed in five different locations (Fig.2) inside the Giant Ice Cave with a mock-up of the WISDOM radar accommodated on three different platforms (Fig.3): (a) the WISDOM wheeled cart, (b) the so called “Cliffbot” developed by the French Planet Mars Association, and (c) the radio-

4

controlled « Magma-White » rover from ABM Space Education in Poland described in Meszyński et al. (2013).

Figure 2 : Map of the Giant Ice cave in Dachstein, Austria showing the sites investigated with WISDOM with details of the acquisition schemes for sites B and D.

Figure 3: The three different platforms used during the ice caves field test in Dachstein: a) The WISDOM cart, b) the Cliffbot and c) the Magma-white rover. The two WISDOM antennas are clearly visible Single acquisitions were performed as accurately as possible each 10 cm to build several fewmeters-long 2D profiles along the rover path (Table 1). Every set of measurements was 5

performed above surfaces covered with ice. Additional tests where the location of the antennas was obtained from stereo camera and provided by ProvisG (Morley et al, 2012) were successfully performed. But the WISDOM data acquired were not exploitable and are not used in the paper.

Table 1 : Test setups at the different sites Location (see Fig.2)

Platform

Height above the surface in cm

Number of soundings

Site A

Cliffbot

8

197

Site B

Wheeled Cart

30

1074

Site C

Wheeled Cart

30

142

One profile of 7 m

Smooth ice

Site D

White Magma Rover

17

102

Five parallel profiles of 1 m One profile of 4 m

Smooth ice Rough ice

data acquired Two parallel profiles of 6 m and 8 m Grid : see acquisition scheme in Fig.2

Characteristic of the surface Smooth ice Smooth ice

Where the cave configuration made it possible (on Site B), additional parallel and perpendicular profiles were performed to aim at a 3D mapping of the subsurface (see Fig. 2 left top). While the nominal distance between soundings (i.e. 10 cm) would allow a correct migration of the 2D profiles (assuming that all reflections are coming from beneath the antennas), the distance between parallel profiles was larger than 0.5 m which makes the 3D migration of the data set impossible. Nevertheless, our purpose is to use the whole set of data to reconstruct as accurately as possible the 3D structure of the subsurface.

3. The Instrument WISDOM is a GPR that operates over a frequency range of 0.5-3 GHz. In contrast with traditional impulse radar, WISDOM is a Step Frequency (SF) radar that transmits a series of continuous wave (CW) signals, each at a single frequency over a time step of duration ∆t (Ciarletti et al., 2011). This step frequency radar system has a lot of advantages compared to the impulsion one in time domain, starting with the maximization of the energy sent for each single frequency, and its stability or repeatability (Ardekani and Lambot, 2013). It also offers 6

the possibility to accurately estimate the subsurface dielectrical properties (Lambot et al., 2003, Lambot et al., 2004).

The radar is designed to receive the echoes concurrently as the transmitter is transmitting, which reduces significantly the blind zone of the instrument. It can also be commanded to delay its reception to focus on deeper sounding (hardware gating). To down-convert the signal, the received signal is mixed with the transmitted signal and then filtered by a 0.6 kHz bandwidth low-pass filter, which allows to reduce the noise level. A complete set of measurements is performed over a number of bandwidth

frequencies that cover the radar frequency

and is referred to as one sounding or single acquisition. The instrument’s

nominal parameter values are presented in Table 2.

Table 2 : Nominal main radar parameters values Central Frequency 1.75 GHz Frequency Bandwidth 2.5 GHz Step duration ∆t 200 µs Number of frequencies 1001 Frequency Step ∆f 2.5 MHz Radiated Power Pe [16-19] dBm

An Inverse Fourier Transform (IFT) allows retrieving the time-domain impulse response from which the propagation delay and amplitude of echoes from the surface, various interfaces and reflectors can be readily determined. When the radar is operated periodically along the rover path (to perform a profile), one can build a 2D radar image, referred to as a radargram, whose intensity depends on surface and interface reflectivity, and on losses encountered during the propagation (see Fig.4), with respect to the horizontal positon of the antennas and the propagation delay of the wave. When the permittivity is known, delays can be converted into distances to build a 2D profile of the subsurface.

7

Radar ability to resolve fine layering is linked to the frequency bandwidth of the signal hence to their center frequency (Yilmaz, 2001). WISDOM’s frequency bandwidth leads to a synthetic pulse total duration of 1.2 ns, which results, for example, in a theoretical range resolution better than 8 centimeters in the ice considering a wave velocity of 168 m.μs-1 (Fletcher, 1970).

Single sounding

Non migrated radargram 15 Direct coupling

20

Surface echo

25

Internal reflections

Delay (ns)

30 35 40

Scaterrers signatures

45 50 55 60 b)

a)

65 0

2 4 Distance (m)

6

Figure 4 : a) Example of a non migrated radargram collected during the Dachstein field test on site B with the wheeled cart. b) Signal collected at distance 5 m is shown as an example. Both the signal (dotted line) and its envelope (solid line) are displayed A typical radargram is shown in Fig.4. For each radargram in the following, the same processing is performed: a windowing is applied to the data collected in the frequency domain, which are converted into time domain using an inverse Fourier transform, and an interpolation by three, both in time and space direction is performed. Eventually, radargrams are normalized. For this particular radargram, no gain was applied. The received signal is displayed for one sounding on the right. The first pulse (black-white-black sequence) 8

corresponding to the direct coupling between the transmitting and receiving antennas (straight line on the radargram) is noticeable at 15 nanoseconds. All echoes arriving before this one come from the electronic of the instrument, and for this paper, we define the time zero to be at the beginning of this direct coupling. The surface echo arrives just beneath, with a duration between 1 nanosecond and 1.5 nanosecond (depending on the surface roughness). A number of hyperbolic-like shapes, which are typical signatures for small reflectors, are visible too. In addition, one can notice reflections inside the ice (top-right of Fig.4), which suggest the WISDOM’s capacity to resolve internal layering. This particular point is discussed in the 6.1 section. The wide frequency band allows to subdivide our data in frequency sub-bands to discriminate contribution from small or large scatterers, while keeping an acceptable resolution. Non-migrated radargrams derived from WISDOM data in site C (see Fig.2) are showed on Fig.5.

All freq a)

10

Travel time (ns)

LF

HF b)

c)

20 30 40 50 0

2 4 6 Distance (m)

8

0

2 4 6 Distance (m)

8

2 4 6 Distance (m)

8

Figure 5: Radargram recorded in site B in polarization 0/0 (see Fig.6). a) Full frequency bandwidth from 0.5 to 3 GHz ; b) Lowest half bandwidth (LF) from 0.5 to 1.5 GHz; c) Highest half bandwidth (HF) from 2GHz to 3 GHz. The radargram on the left of Fig.5 has been generated using the full WISDOM frequency band-width i.e. from 0.5 GHz to 3 GHz, while the ones on the middle and on the right have been obtained respectively with the lowest frequencies (LF) from 0.5 GHz to 1.5 GHz and with the highest frequencies (HF) from 2 GHz to 3 GHz. The same exponential gain was applied for 9

each frequency band. While the lowest frequencies allow sounding deeper into the subsurface, the highest frequencies give access to small structures (but only in the very shallow subsurface) since it interacts with objects whose sizes are inferior to the minimum wavelength of the lowest bandwidth, i.e. objects smaller than 11 centimeters, assuming a velocity in the ice of 168 m.μs-1 (Fletcher, 1970). Anyway, the use of the half bandwidth implies a loss in resolution. The full frequency bandwidth gathers these advantages, hence displaying a much better resolution. Radargrams enhance the transparency of the medium made clear by strong reflections from embedded scatterers, implying very low losses. One can identify a number of hyperbolic shapes corresponding to embedded rocks. WISDOM is a fully-polarimetric GPR. The transmitting and receiving antennas are identical, both composed of two orthogonal Vivaldi elements (Ciarletti et al., 2011) which are in a vertical plane at an angle of + 45° and -45° with respect to the rover direction. The transmitting and receiving antennas are in contact with each other, leading to a space between the two centers of about 10 centimeters. For each single acquisition, two antenna’s elements are selected (one for transmission and the other one for reception) according to the desired polarization (see Fig.6). WISDOM can thus provide a complete two by two scattering matrix by recording both co-polarization (1/1 or 0/0) and cross-polarization (0/1 or 1/0) data. In this paper, we focus on co-polarization measurements. The fact that the radiation pattern of the Vivaldi radiating elements is roughly aligned in the direction perpendicular to the radiating element’s plane rather than isotropic (Plettemeier et al., 2009) can be used to provide additional information about the 3D location of reflectors (see section 5.3).

10

Figure 6 : Schematic view of WISDOM transmitting antenna configuration and footprint on the ground. If the receiving antenna’s radiating element is aligned with the transmitting one, the configuration is in co-polarization mode (1/1 or 0/0). If they are perpendicular, the configuration is in cross-polarization mode (1/0 or 0/1) 4. Retrieval of the subsurface permittivity from the surface echo The permittivity value of the medium controls the velocity of the electromagnetic waves and hence the conversion of measured delays into distances. Moreover, even if the dielectric properties do not allow the unambiguous identification of a material (Guéguen and Palciauskas, 1994), the ability to estimate permittivity values can help to constrain the composition and porosity of a geological unit that is detected on a radargram. Radar data allow getting an estimate of the spatial distribution of the dielectric properties. The relative permittivity defined at a given frequency value

( )

( )

for a continuous wave has a complex

( ) . The real part of the relative permittivity value

is also

called the dielectric constant of the medium. For this paper, given the very low losses expected in ice at WISDOM frequencies (Fujita et al., 2000), the velocity of electromagnetic waves can be simplified to

√

11

.

(1)

Thus the need is to focus on the ways to estimate the velocity of the waves and thus the dielectric constant value

.

It is possible to measure the dielectric constant value of collected samples in laboratory (ElShafie and Heggy, 2013, Cereti et al., 2007, Allen et al., 1998, Heggy et al., 2001, Williams and Greeley, 2004, Pettinelli et al., 2005) but the properties of the sample (mainly the density which is a very important driver of the permittivity value) are altered in the process (Guéguen and Palciauskas, 1994) thus the need to get in-situ estimates of the dielectric constant of the subsurface. WISDOM provides two different complementary methods to do that.

4.1. Dielectric constant estimated from the surface echo The first method is based on the surface echo amplitude and provides an estimate of the top layer’s dielectric constant and its variability along the rover. Since the nominal antennas accommodation on the ExoMars rover at around 37 centimeters above the surface allows disentangling the surface echo from the direct coupling between antennas, we gain access to the amplitude of the surface echo along the rover path (Slob et al., 2010, Chanzy et al., 1996). Our purpose here is to take advantage of this information to obtain an estimate of the dielectric constant of the surface layer.

4.1.1.

In case of a plane and smooth surface

For a smooth plane reflecting surface, the power of the returning echo from the surface can be derived from the Friis radar equation (Friis et al., 1957) as in (2) provided that the reflecting plane is in the far field of the antenna. (|

|√

)

12

(2)

where

is the power of the returning echo from the surface,

coefficient of the returning echo from the surface,

is the reflection

is the transmitted power,

the gain respectively for the transmitting and receiving antennas,

and

are

is the wavelength and

is the distance between the antennas and the surface. For a smooth surface at normal incidence in the far field, the reflection coefficient

only

depends on the dielectric constant of the top layer.

√ √

.

(3)

The other parameters of equation (2) - contributing to the K factor - are linked to the radar settings and geometrical factors; as a consequence, factor K can be measured experimentally using a perfectly conducting plane target (reference) placed at the same distance the radar would see the surface (Slob et al., 2010, Maser and Scullion, 1992). For such a conductive target, the reflection coefficient is known, and the measured power

Eventually, we eliminate the

is

(4)

factor and obtain,

which leads for the amplitude to

|

| .

Thus

√

|

,

(5)

(6)

|.

(7)

It is important to estimate the impact of the experimental conditions on the retrieved dielectric constant value. In fact, the far field is at about 1.5 m in the vacuum (Bansal, 1999) over the full frequency bandwidth, which means that the surface is not in the far field of the antennas. Moreover, the metallic plate we use to measure 13

is of limited dimensions. We have run

simulations to quantify the resulting error of the dielectric constant value when we use relation (7) and check whether it is acceptable for our purpose. These simulations have been performed with the TEMSI-FD (Time ElectroMagnetic SImulator-Finite Difference in time domain) Software Developed in the Xlim Institute (Limoges, France), which allows to model the actual shape and behavior of the WISDOM antennas and its surrounding environment. It was used to simulate the calibration process for the geometry and dimensions of the metal plate used on the field test. We have considered a plate of 50 cm by 100 cm for a typical distance equal of 35 cm, which corresponds to an effective antenna aperture of 2 x 45°. Results show that the magnitude of the reflected field is proportional to the permittivity contrast at the surface. The simulated data show that under our experimental conditions, the error on the estimated dielectric constant value is ~ 0.1 for the ice, considering a typical value of 3.2 ((Fletcher, 1970, Fujita et al., 2000), corresponding to an acceptable error of 3 cm at a depth of 2m. As a consequence, the reference measurement performed with the same radar settings and geometrical configuration will allow us to use equation (9) to retrieve the dielectric constant value provided that the radar is operated on a plane and smooth surface.

4.1.2.

In case of a rough surface

The method described above requires the natural surface to be as flat and smooth as possible, in order to minimize the impact of scattering and multiple reflections at the surface. If the surface is not smooth enough, these phenomena will induce a reduction of the reflected amplitude in the direction of the receiving antenna and will lead to an underestimated value of the subsurface dielectric constant. The purpose here is to have a clear understanding of the roughness impact on the dielectric constant retrieval and to propose a validity domain for our method.

14

WISDOM’s ability to measure the travel time of the scattered wave at the surface gives an indirect access to the surface’s roughness. Actually, we are able to retrieve an image of the variation of the surface’s height along the rover path as sensed at the radar frequency and integrated over the instrument foot print. Given the WISDOM’s broad frequency bandwidth, we can split it into two frequency sub-bands and still have a range resolution good enough to detect the surface. At lower frequencies, we anticipate that the radar waves will interact with the surface as if it were smooth and that equation (9) will give a more accurate estimate of the dielectric constant than at higher frequency. This method has been applied on the data collected in Dachstein where all surfaces were made of ice for which we know the dielectric constant. 4.2 Results Over 1300 co-polar single acquisitions were performed with both the WISDOM cart and the White Magma Rover on very smooth ice and around 100 on a much rougher surface mainly composed of ice debris which size vary from a few millimeters to 2 centimeters (see left of Fig.7) with the White Magma Rover only. The sufficient height of the antennas above the surface allowed to obtain an image of the variation of the surface’s height seen by the radar along the rover path. As previously explained, we consider a lower sub-band from 0.5 GHz to 1.5 GHz. In case of very shallow targets (for example at the surface or just beneath it), it is also possible to define a high frequency sub-band from 2 GHz to 3 GHz. Figure 7 shows an example of the same rough surface perceived by the radar using the full frequency bandwidth (top plot), at low frequency (middle plot) and at high frequency (lower plot).

15

Figure 7: Variation of the surface’s height seen by the radar. The measurements were performed at site D with the Magma rover. a) is obtained with the full frequency bandwidth, b) with the lower frequency sub band and c) with higher frequency sub band. As expected, the surface seen by the instrument is smoother at low frequencies than at high frequencies. The highest frequencies are the most impacted by the surface roughness. To quantify the roughness sensed by the radar, we derive a radar roughness parameter which is the root mean square of the heights seen by the radar. This parameter has the dimension of a distance and is linked to the geometrical roughness of the surface but also to some instrument’s characteristics such as the frequency band and the foot print. In addition to the

obtained with the WISDOM’s full frequency bandwidth, we also compute a

value for the lower and for the higher frequencies. For each platform and each site of investigation, we used equation (9) to estimate the dielectric constant of the ice. The results for the radar roughness parameter and the dielectric constant together with its standard deviation are given in Table 3.

16

Table 3 : Dielectric constant estimates and radar roughness parameter Site B Smooth ice with the WISDOM cart Number of soundings 1074 5∙10-3 m 3.2 0.5–3 GHz -3 1∙10 m 3.4 0.5–1.5 GHz -3 8∙10 m 3.0 2–3 GHz Site C Smooth ice with the WISDOM cart Number of soundings 142 0.5-3 GHz 3∙10-3 m 3.2 -3 0.5-1.5 GHz 3∙10 m 3.0 -3 2-3 GHz 7∙10 m 3.0 Site D Rough ice with the Magma White Rover Number of soundings 102 0.5-3 GHz 6.6∙10-3 m 2.3 -3 0.5-1.5 GHz 2∙10 m 3.3 -2 2-3 GHz 1.1∙10 m 1.8

0.25 0.25 0.30

0.20 0.30 0.50

0.47 0.30 0.30

Figure 8: Estimated dielectric constant for three sub-bandwidths according to the platform and the radar roughness parameter Results in Table 3 and Fig.8 show that, whatever the frequency range we have considered, a value below 3.10-3 m allows a correct retrieval of the dielectric constant value. Even when the surface is rough, the use of the radar’s lowest frequency band allows a correct estimate of the permittivity value. 17

As previously mentioned, this method requires the antennas to be at a minimum distance above the surface so that the direct coupling between the antennas and the surface echo can be disentangled. During our field test, the WISDOM cart with an h value of 30 centimeters provided the best configuration. The White Magma Rover with 17 centimeters still provides the possibility to retrieve the surface echo but the direct coupling between antennas and the surface echo slightly overlap which explain the higher error bar on the estimated dielectric constant value. On the other hand, the data acquired with the Cliffbot platform could not be used because the distance between the antennas and the surface was around 8 cm. Based on experimental tests, 15 cm is the minimal height above the surface that can allow this method to provide a retrieval of the permittivity. The use of the radar surface echo’s amplitude has previously been applied to the MARSIS radar data collected from orbit at a few MHz in order to get an estimate of the Martian subsurface permittivity (Carter et al., 2009a, Carter et al., 2009b, Campbell et al., 2008, Lauro et al., 2010, Lauro et al., 2012, Picardi et al., 2005, Watters et al., 2007, Orosei et al., 2014). In the case of MARSIS, the surface could be considered to be in the far field domain and the roughness of the surface was estimated from the MOLA numerical model of the Martian surface (Picardi et al., 2004). In the case of WISDOM, the much smaller dimensions of the instrument and of the distance to the surface make possible a much more accurate calibration on a target in the same field conditions as it will be for the ExoMars 2018 mission. Such a calibration with a perfectly conductive plate will be performed on Earth before launch after the instrument is integrated on the rover. Upon arrival on Mars, it will not be possible to repeat the same measurement. Nevertheless, the amplitude of the direct coupling between antennas (which is not impacted by the nature of the surface) will allow us to monitor and correct for any change in the transmitted power that might have occurred during the cruise or might be linked to the change in environmental conditions. The possibility to perform a few 18

soundings for calibration purpose on the landing platform before launch and just before the rover egress once on Mars is under study. Even if the landing platform is not a smooth and perfectly conductive target, this would provide a possible comparison on the same target.

5.

Information obtained from the buried scatterers signature

The second method is well known to geophysicists. It is based on the shape of individual scatterer’s typical signature. 5.1 Dielectric constant estimation In case of GPR with antennas in contact with the surface, a small reflecting body embedded in a homogeneous medium will generate on a radargram a mathematically exact hyperbola. This hyperbola’s parameters can be used to retrieve the signal velocity in the medium, thus the distance of the scatterer from the surface and the average dielectric constant of the material above the scatterer (Huisman et al., 2003). In the case of WISDOM, the antennas are at a non-negligible height

above the surface,

which implies that the wave partially propagates in the air and undergo refraction at the interface between air and ground. As a consequence, the signature on the radargram, even if it looks like one, is no more an exact mathematical hyperbola and needs to be resolved numerically. We developed a simplified and hence very fast direct model based on the ray tracing approximation to simulate the propagation in the air, the transmission at the surface and the propagation through a number of homogeneous layers to and from a scatterer of negligible dimensions. We focused on the propagation delay of the reflected wave rather than on their amplitude but the simulations account for the actual antenna aperture for each polarization. First of all, the simulated data show that, as expected, the impact of the antenna height above the surface cannot be neglected especially for shallow targets (Dorizon and Ciarletti, 2013). This code was then run intensively to create a database of simulated 19

signatures for comparison with experimental data. The best least mean square fit to the experimental data provides an estimate of the layer’s dielectric constant, of the horizontal position along the path and of the distance to the scatterer (not to be mistaken with the actual depth of the scatterer). An accurate knowledge of the antennas’ actual location for each acquisition is necessary to retrieve the dielectric constant value. In fact, Monte Carlo simulations show that, even in the case of a perfectly homogeneous layer, a Gaussian random error with a standard deviation of 2 cm on the horizontal position would lead to a relative error around 4% on the estimated distance in the subsurface, i.e. 8 cm for a depth of 2 m. For each signature detected on the radargrams, we have estimated the mean dielectric constant value between the reflector and the antennas. Figure 9 shows the histogram of the obtained values from the data collected at site B. Dielectric constant values equal to 1 correspond to propagation in the air and thus to interaction with surface clutter that need to be eliminated in the sub-surface analysis. Signatures of small scatterers embedded in the ice should give a dielectric constant value close to the water ice one, i.e. 3.2 (Fletcher, 1970). Nevertheless, direct observations reveal that the ice might not be pure (see photo on Fig.11) and that a small fraction of bubbles, liquid water or dust might be present inside (Clausen et al., 2007). As a consequence we assume that values between 3.2 and 3.8 are due to reflections on scatterers that are either embedded in the ice or located at the interface between the ice and the bedrock. All these reflectors are at distance ranging from 1.5 m to 4 m from the antennas. Eventually, values above 3.8 are linked to waves that travel partially or totally in the rock below the ice; they all correspond to distances larger than 4 m and are thus very likely to be beneath the interface ice/bed-rock.

20

Figure 9: Histogram of the dielectric constant values retrieved for the hyperbolic shaped detected on radargrams. 5.2 Migration Once the permittivity value is determined, we can build a velocity model to migrate radargrams. We use the matGPR (Tzanis and Kafetsis, 2004) software to produce migrated radargrams using a velocity model that takes into account the first air layer (0.3 m.ns-1), a layer of 2 meters thick formed of pure ice (0.168 m.ns-1) and a half space made of limestone rocks with a velocity of 0.12 m.ns-1 (Milsom, 2007). Resulting migrated radargram is displayed on the right of Fig.10, together with the corresponding non-migrated radargram. As expected, the shallowest diffracting objects are well constrained (continuous line), but as soon as we reach a propagation delay of about 15 nanoseconds, scatterers are not as well constrained as previously (stippled lines). However, the bed-rock seems to be well restituted. This suggests that the ice is pure in the first meter of the subsurface, but that deeper the ice contains impurities implying an increase of the mean permittivity value.

21

Non migrated 5

Migrated velocity 0.168 m/ns a)

b)

0.5

10 15 1.5

20 25 Depth (m)

Travel time (ns)

1

30 35 40 0

2

4 6 Distance (m)

2 2.5 3 0

8

2

4 6 Distance (m)

8

Figure 10 : Radargrams obtained on site B and produced with the full frequency band in polarization 0/0. a) Non migrated radargram; b) Migrated radargram with a velocity of 0.168m.n-1. The continuous line circles highlight the well constrained features whereas the stippled lines enhance the less resolved objects. The same work was done to migrate radargrams from site D. Scatterer signatures provided a velocity of 0.150 m.ns-1, and results on Fig.11 show that they are well constrained if we consider this velocity for the migration, leading to a permittivity value of 4. This is consistent with the hypothesis of ice containing impurities that can be either homogeneously distributed or located inside layers, which is the most probable hypothesis since layering can be observed directly in numerous places inside the cave. This particular point is discussed in the 6.1 section.

22

Migrated velocity 0.168 m/ns

Non migrated 0

a)

0.5 b) 1 Depth (m)

Travel time (ns)

10 20 30

2

40 50

1.5

2.5 3

3 3.5 4 4.5 Distance (m)

3 3.5 4 4.5 Distance (m)

Figure 11 : Radargrams from site D produced with the full frequency band in polarization 0/0.a) Non migrated radargram; b) Migrated radargram.

5.3 Using co-polarization information to constrain the 3D location of a scatterer As explained in the 5.1 section, for each individual scatterer, we are able to get an estimate of its distance to the antennas. Assuming an error of 0.1 for the dielectric constant retrieval, this leads, through an error propagation method, to an error of about 4 centimeters on the retrieved distance for an object at 2.2 meters. The actual depth cannot be determined at this step but we know that the scatterer is located on an arc of circle in a plane perpendicular to the rover direction and which center is the antenna location. It is nevertheless possible to take advantage of the additional information provided by the amplitude of the scatterer signature to further constrain its 3D location. The shape of the antennas radiation pattern (see Fig. 6) and related footprint on the surface is not symmetrical with respect to the rover direction but oriented at +/- 45° depending on the polarization chosen. Hence, if we consider for example a 23

scatterer located on the right side of the rover track, the amplitude of the wave transmitted by the radar in the scatterer direction will be weaker in polarization 0 when the rover is getting closer than when it is moving away. Means the scatterer on the right side appears first in polarization 1 and later in polarization 0. The difference in distance of the maximum detection for both perpendicular polarizations is a measure for the across track location of the scatterer, and further studies plan to investigate this particular behavior to better constrain the scatterers 3D localization. We can thus expect that the hyperbolic-like signature of this scatterer will show a dissymmetrical level of amplitude with a first leg weaker than the second one when observed in polarization 0/0 (see Fig.12) and that it will be the opposite when the same scatterer is observed in polarization 1/1.

Figure 12 : Schematic view of a configuration in polarization 0 with a scatterer on the right of the rover path. Corresponding hyperbola-like signature This consideration does not take into account the radar cross section of the scatterer itself that could be different for the two considered co-polarizations but just the antenna gain for both the transmission and reception. We nevertheless expect to take advantage of this difference to 24

identify off-track small reflectors and better constrain their location. We have used this technique to determine whether a given scatterer is likely to be on the right side or left side with respect to the rover path or even undecidable when the difference in amplitude is not significant. This method has been previously validated on very simple configurations with spheres hanging in the air or buried in sand. The Dachstein field test provided the opportunity to apply the method on more realistic configuration. Indeed, where the cave was large enough (in site B), we acquired seven eight meters parallel profiles, nine three meters cross profiles and two 9 meters diagonal profiles in order to investigate the whole area (see Fig.2 on site B). For each scatterer detected on each radargram, we have defined, according to the method explained above, a status with respect to the rover path: left, right or undecidable. On the other hand, if the scatterer is detected on more than one radargram, we can constrain its 3D location to the intersection of the corresponding circular arcs. The consistency of the results obtained by both methods is illustrated on Fig.13 for a scatterer that was detected on 4 radargrams. The intersection of the four circular arcs is consistent with the status obtained using the radargrams in co-polarization configuration. This result suggests that even in the case of a single rover path we can decide whether an off-track buried scatterer is on the right side or left side of the rover path.

25

Figure 13 : Retrieval of the 3D location of a scatterer detected on four different rover paths. Each number corresponds to one rover path performed on the grid and is visualized by an arrow. For each number, we have obtained a circular arc for the possible location of the scatterer and determined its status. The arcs intersection provides the best estimate of the scatterers actual 3D location. We assume an error of

for the horizontal positions considering the spatial step of ten

centimeters between acquisitions, leading to an arc width of ten centimeters. A thickness of is considered for each circular arc, according to the error estimated on the dielectric constant and thus the depth. The more the same object is detected on different radargrams, the better its position will be constrained. Eventually, we define the point where the number of intersections is maximal to be the most probable location for the center of the scatterer. However, the size of the object is not taken into account. 5.4 Taking advantage of cross-polarization measurements As explained before, it is possible to take advantage of both co-polarization configurations to better constrain scatterers 3D position. The comparison between co and cross-polarized data is an additional help to determine the scatterer shapes but also the features’ orientations. Fig.14 26

shows two 2D profiles in co and cross-polarizations in site D converted into depth assuming a uniform velocity of 0.168 m. ns-1. One can notice two trapped scatterers at 1.8m depth on copolarized data, which are almost undetectable in cross polarization. This indicates that these two scatterers are roundish, or at least not angular. On the contrary, if objects are detected on both polarizations, it suggests that these rocks and boulders are very angular. This is the case on Fig.14 at the transition between ice and rock material at about three meters depth, implying a much fractured bed-rock or an assembly of angular rocks and boulders.

All freq polarization 0/0

All freq polarization 0/1

1 a)

b)

Depth(m)

1.5 2 2.5 3 3.5 8 9 10 Distance (m)

8 9 10 Distance (m)

Figure 14 : 2D profile from site D using all frequencies. a) Co-polarization 0/0; b) Crosspolarization 0/1 Additional information can be obtained about the geometry of the reflectors when comparing polarizations. An illustration is given in Fig.15, where we can notice, at about one meter depth starting at the half of the profile the detection, on cross-polarized data, of a banded layer that is invisible on co-polarized data. This suggests that this is not due to scattering but rather to a large scale feature which 3D orientation favors a change in the waves’ electric field orientation at reflection. 27

LF, polarization 0/1

LF, polarization 0/0 0.5

a)

b)

1

Depth(m)

1.5 2 2.5 3 3.5 4 4.5 0

2

4 6 Distance (m)

8

0

2

4 6 Distance (m)

8

Figure 15 : 2D profiles from site B generated with the lowest frequencies (LF). a) Co polarization 0/0; b) Cross polarization 0/1 6. Sites characterization in the cave and 3D mapping of the subsurface Many studies have been conducted to study ice cores and determine their condition of formation (Eisen et al., 2003; Eisen et al., 2007; Arcone et al., 1995). Variation in the ice dielectric properties potentially depend on crystal orientation fabric, liquid water inclusions, air bubbles or dust content, ice chemistry, and temperature conditions during permafrost genesis. When diaphanous, cloudy bands layering appear inside the ice, it can be caused by layers of small bubbles (Clausen et al., 2007), or in case of darker and brown layers, it could originate from an accumulation of cryogenic calcites (Obleitner and Spötl, 2011; Clarke and Lauriol, 1992; Haussmann and Behm, 2011) or dust. In each investigated site, the obtained radargrams are in good agreement with the ice characteristics: it appears as rather homogeneous with horizontal or tilted low amplitude reflections as the wave encounters small dielectric properties changes. Strong reflections are clearly visible deeper into the subsurface (see Fig.5), enhancing the transparency of the ice

28

above. Many diffraction signatures due to rocks, boulders or fractured bedrock appear at the transition between ice and bedrock, underneath the ice sheet or trapped inside the ice. 6.1 Ice characterization In opposition to Haussmann and Behm’s conclusions (2011), our data show that the ice in the cave is not homogeneous but layered. A zoomed portion of a radargram converted into depth using a homogeneous wave velocity of 0.168 m.ns-1, is presented on Fig.16 b) together with a photograph Fig. 16 a) of the exposed actual ice’s layering at the same scale. From the photograph, we can identify layers with different tints, from a diaphanous material to a very brown colored layer. These layers have a thickness varying from a few to ten centimeters. The picture and the radar acquisition were not taken in the exact same place, which prevents from performing a quantitative comparison but we can assume that the ice layers have the same order of thickness within the cave.

Figure 16: a) Picture of exposed fine layering inside the ice layer taken along the path between Tristandom and site A (see Fig.2). A 2 euros coin (diameter 2.56 cm) provides the scale. b) Zoomed radargram showing the layering of the ice and the transition with the bed29

rock in site B. The depth is obtained assuming a uniform dielectric constant wave velocity of 168 m.μs-1. The main question about these layers detected by the radar is whether they are due to instrumental artefacts, multiples or actual layers. Instrumental artefacts usually give parallel features that are easily detectable. Following Haussmann and Behm analysis (2011), we can also consider that wherever a diffracting signature can be found at the termination of a layer, the feature on the radargram comes from a real interface. Given the WISDOM frequency bandwidth, the theoretical resolution (ability to discriminate between two close echoes) is around 7 cm in water ice. We anticipate that the achieved resolution can be increased when interpreting from a radargram: layers can be followed along the antennas path, even if some contributions may slightly overlap on a single acquisition. From all these observations, we can conclude that the ice volume is formed of fine strata with different electromagnetic properties, i.e. different compositions, such as calcite powder inclusions or trapped air bubbles. 6.2 Sites characterization The dielectric constant allows us to convert our radargrams into depth to produce 2D profiles. Five different sites were investigated during the Dachstein Mars Simulation. The use of copolarized data with the full frequency bandwidth gives a first overview of the subsurface geometry. Comparison in sub-bandwidths and between co-polarized and cross-polarized data are additional help to constrain the shape and features orientations in the subsurface or the sizes and shapes of embedded rocks or boulders. Results regarding the geometry in the different sites inside the Giant Ice Cave are summarized in Table 4:

30

Table 4 : Sites characterization from 2D profiles

Site

Corresponding 2D profile

Ice sheet cover

Bed-rock

Additional information from comparative study in frequency sub-bands or polarizations

LF polarization 0/0

1

Site A

Depth(m)

2

Angular rocks and boulders Transition Many rocks and boulders whose sizes vary between between ice trapped inside the ice. Almost less than ten centimeters to and bed-rock no layering detected one or two meters based on is not clear sub frequency band analysis

3

4

5

6 0

2

4 Distance (m)

6

All freq polarization 0/0

Site C

Depth(m)

0.5

Much fractured bed-rock or Layering detected. A first layer an assembly of angular inside the ice of about 50cm Bed-rock boulders whose sizes vary thick, followed by thinner between 50cm from a dozen centimeters to layers before reaching the bed- and 1m depth. a few meters based on sub rock frequency band analysis

1

1.5 0

2

4 Distance (m)

6

All freq polarization 0/0 1

Site D

Depth(m)

1.5

Two trapped scatterers at 1.8m depth, no fine layering detected except for one clear tilted layer between 1.2m and 1.5m depth that could originate from calcite powder inclusion

2

2.5

3

8

8.5 9 Distance (m)

9.5

10

31

Scatterers’signatures at 1.8m depth suggest they are roundish, based on comparison between Bed-rock polarization modes, and between 2.5m have dimensions superior to and 3m depth 15cm, based on subfrequency band analysis. The bed-rock is very likely to be much fractured

Measurements for site B are discussed in the following section. 6.3 3D localization of scatterers Based on the method described in the 5.3 section, we have reconstructed a 3D map which represents the most probable locations of the main scatterers. All profiles in both co polarizations were used to produce this reconstruction. The reflectors located close to the edges of the investigated area hardly benefit from the method and their 3D location is poorly estimated. For each individual reflector, the dielectric constant value corresponding to the mean value seen by the wave is also obtained. A mean dielectric constant value equal to 1 (detected nine times as shown on Fig.9) allows the clear identification of reflectors that are located above the surface and that must hence been removed from the subsurface mapping process. 3D locations of the scatterers obtained with this method are presented in Fig.17. We have mapped the reflectors’ locations corresponding to a mean dielectric constant value between 3 and 3.4 – consistent with water ice. As mentioned before, it means that the material above the scatterer has the same mean value as for ice, meaning ice is at least reaching this depth. Scatterers are then located at the interface between ice and rock, individually embedded in ice, or embedded in an icy material with a higher rate of dust or liquid water. They are all located at a depth ranging from 1.5 m to 2.5 m.

32

Figure 17 : 3D reconstruction of the bedrock at site B in co polarized (left) and cross polarized (right) configurations. The dashed lines between P1 and P2 represent the rover path that allowed building the 2D profile displayed in Fig.18 The 3D representation shows that the transition between ice and rock material takes place at a depth between 1.50 meters and 3 meters. The ice layer is thus approximatively 1.8 m thick and is almost free from embedded rocks. The comparison between the processed radargrams and the 3D reconstruction of the bedrock allows us to comfort this result.

P1

P2

Individual scatterers

Ice/Bedrock formed of

Debris, rocks Fractured bedrock

Figure 18 : 2D profile obtained when the rover travelled between the points P1 and P2 defined in Fig.17

33

The 2D profile, with a noticeable dip around horizontal positions between 4 meters and 7 meters, and the 3D shape of the bedrock are in good agreement. It has to be noticed that the interface with the bedrock appears to be very rough and even chaotic, and that the number of hyperbolae located below this interface indicate that the bedrock itself is inhomogeneous. This can be interpreted as a bed-rock formed by an aggregation of numerous individual rocks, or as a fractured bed-rock, since fractures and faults in hard rock have similar hyperbolic signatures. This must be distinguished from the typical smooth and easily recognizable layering in case of sedimentation processes. The individual scatterers embedded in the ice above the bedrock visible on the radargram have also been located in the 3D environment. Nevertheless, discrepancies arise from the fact that a radargrams is not a mere slice into the 3D reality but rather a distorted projection on a plane. The low values previously determined just below the surface could be due to a high amount of trapped bubbles of air, since the material is not compacted, whereas the higher value for deeper material could originate from some compacted layers with a larger proportion of impurities or liquid water. 6.4 Comparison with previous studies Although our measurements did not took place at the exact same locations, and the fact that our instrument has very different characteristics from the shielded 200MHz and 500MHz antennas, comparisons between our data and previous studies (Behm and Haussmann, 2008, Haussmann and Behm, 2011) show good agreement. As expected, we observe an ice sheet cover whose thickness varies from a few dozens of centimeters in the tunnel (site C) to three meters at site E in Parsivaldom. The fact that Behm and Haussmann estimated a maximum ice thickness in Tristandom of around 15 meters cannot be confirmed or infirmed by our study, since no measurements were taken in this particular area with the WISDOM GPR. Strong reflections appear at the transition between ice and bedrock, suggesting a much fractured 34

material (limestone) or blocks whose size vary from a few tens of centimeters to one or two meters. We also find rocks totally trapped in the ice, which is consistent with previous results in Tristandom (Behm and Haussmann, 2008). The main difference lies in the internal layering of the ice: Haussmann and Behm concluded that no internal layering was detectable in Rieseneishöhle, whereas we manage to unambiguously detect some layers of about ten centimeters thick. This is also confirmed by little diffraction figures at the end of these layers. Globally, the depths we estimated in the Giant Ice cave in March 2012 are slightly lower than previous studies in 2006 (Behm and Haussmann, 2008). Even if we cannot compare measurements that were not taken in the exact same place, this could be an additional clue to support the thesis of global ice loss tendency (Behm et al., 2009) of this past decades, also mentioned in Haussmann and Behm’s study (2011). We are though unable to know whether this disappearance was progressive or the consequence of some punctual warm events. Conclusions The Dachstein ice caves field test allowed to confirm, in a well-documented environment, the performance of the WISDOM GPR designed for the ExoMars Rover mission in terms of range resolution, ability to provide reliable estimates of the dielectric constant value, the presence and shape of embedded reflectors, and to map a buried interface in a 3D environment. The resolution range we were able to reach in the ice is better than 8 cm for a theoretical value of 7 cm. We have shown that the amplitude of the surface echo at low frequencies can be used to retrieve the dielectric constant value of the near subsurface and have potentially access to its spatial variations along the rover path even in the case of a slightly rough surface. The analysis of the hyperbola-like reflectors’ signatures has confirmed the permittivity values obtained by the first method and allowed the identification and elimination of the reflectors above the surface. Finally, taking advantage of several parallel and perpendicular profiles 35

performed in the same area and the WISDOM ability to perform acquisitions in different polarizations, we have been able to produce a 3D map of the ice/bedrock interface and to locate a few embedded objects localization. If we consider the structure of the ice medium, the sensitivity and the broad frequency bandwidth of WISDOM allow us to observe fine layering in some locations which are confirmed by direct observations performed on exposed ice. This confirms that the ice originates from successive water freezing deposit sedimentation, occurring over decades, even centuries, and could even constitute a record of paleoclimate changes, as explained by Haussmann and Behm (2011) and May et al. (2011). The methods that have been presented and validated in this paper will be applied on data collected in lithic environment namely on Mount Etna and in Atacama desert during the SAFER experiment (Gunes-Lasnet et al., 2014, Dorizon et al., 2014) Even if the pre-selected sites for ExoMars 2018 landing are lithic environments rather than icy ones such as Dachstein, these results demonstrates that the performances of WISDOM are adapted to the objectives of the ExoMars rover mission i.e. characterize the upper 2-3 meters of the Martian subsurface, understand the geological context and identify interesting places to collect samples. WISDOM and the neutron spectrometer ADRON (Mitrofanov et al., 2014) are the only instruments onboard the mission rover that enable to provide information about the subsurface prior to drilling.. Acknowledgments Thanks go to the French CNES and the German DLR for funding resources. We also thank the WISDOM team, i.e. the engineers R. Hassen-Kodja, O. Humeau, B. Lustrement and A. Galic from LATMOS, France, Dr. S.M. Clifford from LPI, Houston, US, for the field test campaign. Finally, our thanks go to the Austrian Space Forum for the organization and access to the caves, and all the personnel that participated in the Mars simulation in 2012.

36

References Allen, C.C. et al., 1998, Martian soil simulant available for scientific, educational study. EOS Transactions, American Geophysical Union 79, 405-409. Arcone, S.A., Lawson, D.E., & Delaney, A.J., 1995, Short-pulse radar wavelet recovery and resolution of dielectric contrasts within englacial and basal ice of Matanuska Glacier, Alaska, USA. Journal of Glaciology, 41(137), 68-86. Arcone, S.A., Lawson, D.E., Delaney, A.J., Strasser, J.C., & Strasser, J.D., 1998, Groundpenetratinng radar reflection profiling of groundwater and bedrock in an area of discontinuous permafrost. Geophysics, 63(5), 1573-1584. Ardekani, M. and Lambot, S., 2014, Full-wave calibration of time- and frequency-domain ground-penetrating radar in far-field conditions. IEEE Transactions on Geoscience and Remote Sensing, vol. 52, no. 1, pp. 664-678. Available from: 10.1109/TGRS.2013.2243458. Bansal, R, 1999, The far-field: How far is far enough?. Applied Microwave & Wireless, 11(11), 59-60. http://people.eecs.ku.edu/~callen/725/Bansal1999AMWpp58.pdf. Barnes, D., Battistelli, E., Bertrand, R., Butera, F., Chatila, R., Del Biancio, A., Draper, C., Ellery, A., Gelmi, R., Ingrand, F., Koeck, C., Lacroix, S., Lamon, P., Lee, C., Magnani, P., Patel, N., Pompei, C., Re, E., Richter, L., Rowe, M., Siegwart, R., Slade, R., Smith, M.F., Terrien, G., Wall, Ward, R., Waugh, L. and Woods, M., 2006, The ExoMars rover and Pasteur payload phase: a study: an approach to experimental astrobiology. International Journal of Astrobiology 5, 221–241. Behm, M. and Hausmann, H., 2008, Determination of ice thicknesses in Alpine caves using géoradar. Proceedings of the 3rd International Workshop on Ice.

37

Behm, M., Dittes, V., Greilinger, R., Hartmann, H., Plan, L., and Sulzbacher, D., 2009, Decline of cave ice – a case study from the Austrian Alps (Europe) based on 416 years of observation. Proc. 15th Intern. Congr Speleol., Kerrville, Texas, 19–26 July, 3, 1413–1416. Bousquet, P.W., Gaboriaud, A., Gaudon, P., Laudet, P., Chiavassa, F., Rocard, F., 2012, French instruments for in-situ missions: Past, present and future. Acta Astronautica 81, 358– 368. Campbell, B.A., Carter, L.M. , Phillips, R.J. , Putzig, N.E., Plaut, J.J., Safaeinili, A., Seu, R., Biccari, D., Egan, A., and Orosei, R., 2008, SHARAD radar sounding of the Vastitas Borealis Formation in Amazonis Planitia. Journal of Geophysical Research, 113, E12010. doi: 10.1029/2008JE003177. Carter, L.M, Campbell, B.A, Holt, J.W, Phillips, R.J, Putzig,, N.E, Mattei, S., Seu, R., Okubo, C.H, and Egan, A.F., 2009b, Dielectric properties of lava flows west of Ascraeus Mons, Mars. Geophysical Research Letters, Vol.36, L23204. doi:10.1029/2009GL041234. Carter, L.M., Campbell B.A., Watters, T.R., Phillips, R.J., Putzig, N.E, Safaeinili, A., Plaut, J.J., Okubo, C.H., Egan, A.F., Seu, R., Biccari, D., Orosei, R., 2009a, Shallow radar (SHARAD) sounding observations of the Medusae Fossae Formation, Mars. Icarus 199, 295– 302. Cereti, A., Vannaroni, G., Del Vento, D., Pettinelli, E., 2007, Electromagnetic measurements on Martian soil analogs: Implications for MARSIS and SHARAD radars in detecting subsoil water. Planetary and Space Science 55, 193–202. Chanzy, A., Tarussov, A., Bonn, F., & Judge, A., 1996, Soil water content determination using a digital ground-penetrating radar. Soil Science Society of America Journal, 60(5), 1318-1326. 38

Ciarletti V., Corbel, C., Plettemeier, D., Caïs, P., Clifford, S.M. and Hamran, S.E., 2011, WISDOM GPR Designed for Shallow and High-Resolution Sounding of the Martian Subsurface. Proceedingsof the IEEE, Vol. 99, No. 5, May 2011. Clarke, I. D. and Lauriol, B., 1992, Kinematic enrichment of stable isotopes in cryogenic calcites. Chem. Geol. (Isotope Geoscie. Sect.), 102, 217–228. Clausen, H.B., Vrana, K., Hansen, S.B., Larsen, L.B., Baker, J., Siggaard-Andersen, M.L., Sjolte, J., and Lundholm, S.C., 2007, Continental ice body in Dobsina Ice Cave (Slovakia) – Part II. – Results of chemical and isotopic study. Proceedings of the 2nd International Workshop on Ice Caves, edited by: Zelinka, J., Demanovska Dolina, Slovak Republic, 29–37. Dorizon, S., & Ciarletti, V., 2013, Subsurface properties retrieval from the ExoMars WISDOM GPR data. European Planetary Science Congress 2013, held 8-13 September in London, UK. Online at: http://meetings. copernicus. org/epsc2013, id. EPSC2013-434 (Vol. 8, p. 434). Dorizon, S., Ciarletti, V., Vieau, A-J., Plettemeier, D., Benedix, W.S., Mütze, M, HassenKodja, R., and Humeau, O., 2014, WISDOM GPR subsurface investigations in the Atacama desert during the SAFER rover operation simulation, communication (oral) during the European Geophysical Union, Geophysical Research Abstracts, Vol. 16, EGU2014-16078-1, EGU General Assembly 2014. Eisen, O., Hamann, I., Kipfstuhl, S., Steinhage, D., and Wilhelms, F., 2007, Direct evidence for continuous radar reflector originating from changes in crystal-orientation fabric. The Cryosphere, 1, 1–10, www.the-cryosphere.net/1/1/2007.

39

Eisen, O., Nixdorf, U., Keck, L. and Wagenbach, D., 2003, Alpine ice cores and ground penetrating radar: combined investigations for glaciological and climatic interpretations of a cold Alpine ice body, Tellus B, 55(5), 1007-1017. ElShafie, A., Heggy, E., 2013, Dielectric and hardness measurements of planetary analog rocks in support of in-situ subsurface sampling. Planetary and Space Science 86, 150-154. Fletcher, N.H., 1970, The Chemical Physics of Ice. Cambridge University Press, June 2009, doi: http://dx.doi.org/10.1017/CBO9780511735639. Friis, H.T., Crawford, A.B., Hogg, D.C., 1957, A Reflection Theory for Propagation Beyond the Horizon. Bell System Technical Journal 36, Issue 3, 627–644, May 1957. doi: 10.1002/j.1538-7305.1957.tb03856.x. Fujita, S., Matsuoka, T., Ishida, T., Matsuoka, K., & Mae, S., 2000, A summary of the complex dielectric permittivity of ice in the megahertz range and its applications for radar sounding of polar ice sheets. Physics of Ice Core Records, 185-212. Gogineni, S., Tammana, D., Braaten, D., Leuschen, C., Akinsg, T., Legarsky, J., Kanagaratnam, P., Stiles, J., Allen, C. and Jezek, K., 2001, Coherent radar ice thickness measurements over the Greenland ice sheet. Journal of Geophysical Research, Vol.106, no D24, 33,761-33,772, December 27, 2001. Groemer, G., Luger, U., Sander, K., Plettemeier, D., 2012, Dachstein Mars Simulation 2012 Mission_Report, https://www.researchgate.net/publication/235753475_Dachstein_Mars_Simulation_2012Miss ion_Report. Guéguen, Y. and Palciauskas, 1994, Introduction to the Physics of Rocks. Princeton University Press. 40

Gunes-Lasnet, S., Kisidi, A., van Winnendael, M., Josset, J. L., Ciarletti, V., Barnes, D. and Chong Diaz, G.,2014, SAFER: The promising results of the Mars mission simulation campaign in Atacama, Chile. I-SAIRAS 2014: 12th International Symposium on Artificial Intelligence, Robotics and Automation in Space, 17-19 June 2014, Montreal, Canada. Hausmann, H. and Behm, M., 2011, Imaging the structure of cave ice by GPR. The Cryosphere, 5, 329–340. Heggy, E., Paillou, P., Ruffie, G., Malezieux, J. M., Costard, F., Grandjean, G., 2001, On Water Detection in the Martian Subsurface Using Sounding Radar. Icarus 154, 244–257. doi:10.1006/icar.2001.6717. Huisman, J.A., Hubbard, S.S., Redman, J.D. and Annan, A.P., 2003, Measuring Soil Water Content with Ground Penetrating Radar : A Review. Vadose Zone Journal 2:476–491. Lambot, S., Slob, E., van den Bosh, I., Stockbroeckx, B., and Vanclooster, M. 2004, Modeling of ground penetrating radar for accurate characterization of subsurface electric properties. IEEE on Geoscience and Remote Sensing, 42(11) :2555 – 2568. Lambot, S., Slob, E.C., Van den Bosch, I., Stockbroeckx, B. Scheers, B., and Vanclooster, M., 2003, GPR design and modeling for identifying the shallow subsurface dielectric properties. Proc. 2nd Int. Workshop Advanced Ground Penetrating Radar, A. Yarovoy, Ed., Delft, The Netherlands, 2003, pp. 130–135. Lauro, S.E, Mattei, E., Soldovieri, F., Pettinelli, E., Orosei, R., Vannaroni, G., 2012, Dielectric constant estimation of the uppermost Basal Unit layer in the martian Boreales Scopuli region; Icarus 219 (2012) 458-467.

41

Lauro, S.E., E. Mattei, E. Pettinelli, F. Soldovieri, R. Orosei, M. Cartacci, A. Cicchetti, R. Noschese, and S. Giuppi , 2010, Permittivity estimation of layers beneath the northern polar layered deposits, Mar. Geophys. Res. Lett., 37, L14201, doi: 10.1029/2010GL043015. Maser, K., and Scullion, T., 1992, Automated pavement subsurface profiling using radar: Case studies of four experimental field sites. Transportation Research Record, 1344, 148–154. May, B., Spotl, C., Wagenbach, D., Dublyansky, Y. and Liebl, J., 2011, First investigations of an ice core from Eisriesenwelt cave (Austria). The Cryosphere, 5, 81–93, 2011. doi:10.5194/tc-5-81-2011. Meszyński, S. and Józefowicz, M., 2013, Analog Mars Rover Service as a Robotic Hardware and Team Building Platform. GSTF Journal of Engineering Technology (JET), Vol. 2 No. 3. Milsom, J., 2007, Field geophysics (Vol. 25). John Wiley and Sons. Mitrofanov, I. G., Litvak, M.L., Sanin, A.B., Starr, R.D., Lisov, D.I., Kuzmin, R.O., Behar, A., Boynton, W.V., Hardgrove, C., Harshman, K., Jun, I., Milliken, R.E., Mischna, M.A., Moersch, J.E., and Tate, C.G., 2014, Water and chlorine content in the Martian soil along the first 1900 m of the Curiosity rover traverse as estimated by the DAN instrument. Journal of Geophysical Research Planets, 119, Issue 7, 1579–1596. doi:10.1002/2013JE004553. Morley, J. G., Lin, N., Muller, J. P., Shin, D., & Paar, G., 2012, PRoGIS: A Web Tool to Understand and pProcess Mars Rover Imagery in a Planetary Context. Lunar and Planetary Science Conference (Vol. 43, p. 2896). Obleitner, F. and Spötl, C.,2011, The mass and energy balance of ice within the Eisriesenwelt cave,

Austria.

The

Cryosphere,

5,

245–257,

discuss.net/5/245/2011/.

42

2011.

http://www.the-cryosphere-

Orosei, R., Jordan, R. L., Morgan, D. D., Cartacci, M., Cicchetti, A., Duru, F., ... & Picardi, G., 2014. Mars Advanced Radar for Subsurface and Ionospheric Sounding (MARSIS) after nine

years

of

operation:

A

summary.

Planetary

and

Space

Science.

doi:10.1016/j.pss.2014.07.010. Pettinelli, E., Vannaroni, G., Cereti, A., Pisani, A.R., Paolucci, F., Del Vento, D., Dolfi, D., Riccioli, S., Bella, F., 2005, Laboratory investigations into the electro-magnetic properties of magnetite/silica mixtures as Martian soil simulants. J. Geophys. Res. 110, E04013. Picardi, G., Biccari, D., Seu, R., Marinangeli, L., Johnson, W.T.K., Jordan, R.L., Plaut, J., Safaenili, Gurnett, D.A., Ori, G.G., Orosei, R., Calabrese, D., Zampolini, E., 2004, Performance and surface scattering models for the Mars Advanced Radar for Subsurface and Ionosphere Sounding (MARSIS) . Planetary and Space Science 52, 149 – 156. Picardi, G., Plaut, J.J., Biccari, D., Bombaci, O., Calabrese, D., Cartacci, M., Cicchetti, A., Clifford, S.M., Edenhofer, P., Farrell, W.M., Federico, C., Frigeri, A., Gurnett, D.A., Hagfors, T., Heggy, E., Herique, A., Huff, R.L., Ivanov, A.B., Johnson, W.T.K., Jordan, R.L., Kirchner, D.L., Kofman, W., Leuschen, C.J., Nielsen, E., Orosei, R., Pettinelli, E., Phillips, R.J., Plettemeier, D., Safaeinili, A., Seu, R., Stofan, E.R., Vannaroni, G., Watters, T.R., Zampolini, E., 2005, Radar Soundings of the Subsurface of Mars. Science, Vol.310, 1925. doi: 10.1126/science.1122165. Plettemeier D., Ciarletti V., Hamran S.-E., Corbel C., Cais P., Benedix W.-S., Wolf K., Linke S., Roeddecke S., 2009, Full Polarimetric GPR Antenna System Aboard the ExoMars Rover. 2009 IEEE Radar Conference, Pasadena, USA (2009). http://hal.archives-ouvertes.fr/hal00439012/fr/.

43

Plewes, L.A., Hubbard, B. 2001, A review of the use of radio-echo sounding in glaciology. Progress

in

Physical

Geography,

June

2001

vol.

25

no.

2

203-236.

doi:10.1177/030913330102500203. Slob, E., Sato, M., and Olhoeft, G., 2010, Surface and borehole ground-penetrating-radar developments. Geophysics, Vol.75, no 5_September_October 2010, P. 75A103–75A120, 12 FIGS. doi: 10.1190/1.3480619. Tzanis A. and Kafetsis G., 2004, A freeware package for the analysis and interpretation of common-offset Ground Probing Radar data, based on general purpose computing engines. Bulletin of the Geological Society of Greece vol. XXXVI, 2004, Proceedings of the 10th International Congress, Thessaloniki, April 2004. Watters, T.R, Campbell, B., Carter, L., Leuschen, C.J., Plaut , J.J., Picardi, G., Orosei, R., Safaeinili, A., Clifford, S.M, Farrell, W.M., Ivanov, A.B., Phillips, R.J., and Stofan, E.R., 2007, Radar Sounding of the Medusae Fossae Formation Mars: Equatorial Ice or Dry, LowDensity Deposits?. Science, New Series, Vol. 318, No.5853 (Nov. 16, 2007), pp. 1125-1128. Williams, K.K., Greeley, R., 2004, Measurements of dielectric loss factors due to Martian dust analog. Journal of Geophysics Research 109, E10006. Yilmaz, Ö., 2001, Seismic data analysis. 2nd edition, Society of Exploration.

44

Highlights :

WISDOM helped characterize the ice thickness and the geometry of the subsurface in different sites inside the Giant Ice Cave. Two techniques are used to retrieve the dielectric constant at the surface and in the subsurface. Roughness has a non negligible impact on the dielectric constant retrieval. An analysis in frequency sub-bandwidths shows that the lowest frequencies are the less impacted by this phenomenon and can be used to estimate correctly the permittivity value. Taking advantage of the polarization mode allows constraining the 3D position of scatterers Fine layering inside the ice is detected.

45