Detectability of subsurface interfaces in lunar maria by the LRS/SELENE sounding radar: Influence of mineralogical composition

Authors


Abstract

[1] The Lunar Radar Sounder (LRS) of the SELENE mission has detected horizontal subsurface features at depths of a few hundreds of meters within all major lunar maria. We have mapped these features at global scale and found a heterogeneous geographical distribution, which correlates negatively with the maps of TiO2 and FeO obtained from UV-VIS measurements by the Clementine probe. High concentration of ilmenite (FeTiO3) in the basaltic mare lava can explain this correlation, this mineral being a strong absorber for low frequency electromagnetic waves. Thus, absence of detection of subsurface interfaces by LRS on a large portion of lunar maria does not necessarily imply that these interfaces are actually absent.

1. Introduction

[2] Lunar maria cover one third of the near-side of the Moon. They consist of more or less circular basins with diameters of a few hundreds of kilometers and depth of a few kilometers. Whereas highlands are ancient and mainly composed of anorthosite formed by crystallization of a primordial magmatic ocean, mare materials are younger and consist of successive solidified lava flows filling pre-existing topographic depressions.

[3] Information on the structure, nature and origin of lunar maria essentially comes from geomorphologic analyses as well as compositional studies using both orbital remote-sensing and laboratory characterization of materials collected during the Apollo and Luna missions. While optical methods only probe the very surface of the Moon (<1 mm), high frequency imaging radars give access to the shallow subsurface (a few meters to tens of meters [Campbell and Hawke, 2005; Thompson et al., 2006]). Results from all these methods indicate that the lunar maria are filled by strata of volcanic layers with basaltic compositions. Crystallization ages range between about 1 Gy and 4 Gy [Hiesinger et al., 2008] revealing a long history of magmatic activity on the Moon. Mineralogy and chemistry of the basalts show significant temporal evolution, characterized by an increased concentration of the mineral ilmenite (FeTiO3) with time [e.g., Staid and Pieters, 2001]. Differentiation of the magma by partial melting and crystallization is responsible for such an evolution. A regolith composed of finely divided mare material produced by the combined actions of solar wind and meteorite bombardment entirely covers the basaltic layers.

[4] Direct probing of planetary subsurfaces to about kilometer depths is made possible by the use of low frequency sounding radar. This method was used for the first time to characterize the structure of the lunar subsurface during the Apollo 17 mission in 1972 (ALSE experiment [Porcello et al., 1974]). Data acquired during two equatorial orbits revealed a clear layering of lava fills in two lunar maria: Mare Imbrium and Mare Serenitatis [Peeples et al., 1978]. Thirty-six years later, the LRS (Lunar Radar Sounder) instrument on-board the Kaguya spacecraft (JAXA/SELENE mission) completed a global coverage of the Lunar surface and subsurface using a similar method, allowing the mapping and characterization of subsurface structure inside lunar maria [Ono et al., 2009]. Subsurface features observed by the ALSE and then LRS radar are interpreted as horizontal interfaces between superposed lava flows of different ages [Peeples et al., 1978; Ono et al., 2009]. Formation of a superficial regolith between successive lava flows has been proposed to produce the contrast of dielectric constant. Oshigami et al. [2009] further discuss this hypothesis. The authors interpret the spatial distribution of subsurface features observed by LRS as a result of variations of the thickness of buried regolith layers. This interpretation is based on the observation that subsurface features are only apparent in the most ancient parts of lunar maria.

[5] The aim of the current study is to build a global map of the distribution of the subsurface interfaces using the LRS dataset. This map is then interpreted at global and regional scales by comparison with surface maps obtained by other remote-sensing techniques.

2. Data and Methods

2.1. Subsurface Dataset

[6] A set of about 300 processed LRS observations or “radargrams”, covering the entire surface of the Moon, was used for this study. The vertical resolution of the radargrams in vacuum is: r0 = 75 m. The actual vertical resolution is: r = equation image where ɛ′ is the real part of the dielectric constant of materials. Assuming plausible values of ɛ′ between 4 and 9 for typical materials of the Lunar upper crust (regolith and dense basalt layer respectively), one can estimate a vertical resolution ranging between 25 and 40 meters. The typical spatial resolution along track is of 75 m per pixel. We stacked the data by averaging 30 adjacent pulses to increase the signal/noise ratio while decreasing the spatial resolution to about 2 km per pixel.

2.2. Surface Dataset

[7] We used the USGS “Lunar Airbrushed Shadedrelief Warped to ULCN2005” digitized map at 64 pixels per degree (ppd) as a reference geographic background for this study, allowing the localization of LRS observations and the direct identification of surface features on radargrams.

[8] The compositional maps of the Lunar surface derived from Clementine UV-VIS observations were also extensively used in this study. Of particular interest for radar analyses are the maps of iron and titanium oxides established by Lucey et al. [2000] and made publically available in a numerical format through the USGS Pigwad server.

[9] In addition to global studies, we carried out a more detailed study of Mare Imbrium. For this purpose, we used the geologic map published by Bugiolacchi and Guest [2008], manually re-projected to fit the shaded relief map.

2.3. Analysis of Subsurface LRS Profiles

[10] In a first step, each LRS profile was visually analyzed to identify potential subsurface features. In a second step, a semi-automatic algorithm designed to map subsurface echoes was used to extract the three-dimensional coordinates (latitude, longitude and depth) of each pixel of a given subsurface feature. This algorithm uses a detection of the maximal returned echo on a given range of distance and depth defined by the user. It allows an efficient estimate of the geometry of subsurface features along with a systematic visual checking of its accuracy by the user. Clutter generated by surface off-nadir echoes is likely to produce signals at an apparent depth below the surface that may be confused with actual subsurface features. However, the large-scale continuity of the observed subsurface features indicates that they are unlikely to be caused by surface clutter. This was confirmed in the case of Mare Serenitatis by physical simulation of surface clutter [Nouvel et al., 2004; Mouginot, 2008] from a numerical Digital Elevation Model (DEM) of the surface [LALT Team, 2009].

3. Results and Discussion

[11] Figure 1 shows examples of the three main types of radargrams that were found in the studied dataset. They differ by the number of interfaces observed in the subsurface: 0, 1 or 2. Assuming an average dielectric constant of 7 for the top kilometer of the Lunar crust, we calculate an average depth of 247 ± 127 m for the shallowest interface and of 488 ± 210 m for the deepest interface.

Figure 1.

Examples of LRS sounding profiles in lunar maria with (a) no visible subsurface interface, (b) a unique shallow subsurface interface and (c) two distinct subsurface interfaces. Assuming a value of the dielectric constant of 7, the shallowest interface is found at an average depth of about 247 ± 127 m and the deepest one at an average depth of 488 ± 210 m. Strong and large hyperbolic features visible on the three profiles are due to off-nadir surface echoes (surface clutter), generally produced by isolated craters. These three examples are representative of most of the profiles measured in lunar maria.

[12] Figure 2 shows vertical profiles of returned echo power versus apparent depth below the surface for the same three examples shown in Figure 1. The strong echo arising from the radar wave reflection on the surface is the most preeminent signal in each case, with amplitude of 50 dB. Following this surface echo, one observes a continuous decrease of the signal due to surface or/and volume scattering of the radar wave. The origin of this scattering, surface or volume, is still unclear and will be the subject of future investigations. The signals due to subsurface interfaces appear as secondary peaks with amplitudes of a few dB and local maxima reaching 10 dB.

Figure 2.

Examples of diagrams showing the evolution of backscattered power versus apparent depth (“A-scope” profiles). The three diagrams are extracted from the corresponding sounding profiles (Figure 1) at places where the subsurface features are the most prominent. Reflection on the surface at nadir is always responsible for the strongest echo. The slow decrease of backscatter energy after the surface echo is due to surface and/or volume scattering and limits the detectability of discrete subsurface features. Echoes returned by the subsurface interfaces usually have amplitudes of a few dB with observed maxima of about 10 dB.

[13] The geographic distribution of the occurrences of the “one subsurface interface” and “two subsurface interfaces” radargrams is represented at global scale in Figure 3. We don't represent the variations of the depth of the interfaces, as the conversion of radar time delay to actual depth requires hypothesis on the value of the dielectric constant that prevents any direct interpretation. Only a limited part of the lunar maria displays the already described subsurface interfaces. After trying to correlate various maps to this distribution of interfaces seen by LRS, it appeared that the concentration map of TiO2 in the regolith determined from Clementine VIS spectroscopy gives the best match. This is evident both at global scale (Figure 3) and at regional scale: the example of Mare Imbrium is presented on Figures 4a and 4b. In addition, comparison with the geological map built by Bugiolacchi and Guest [2008] from geomorphologic and compositional information shows that the distinction between areas showing one unique subsurface interface and areas showing two distinct interfaces corresponds to a putative distinction between two different lower Imbrian series with very similar TiO2 concentrations (1.4 and 1.5 wt. %).

Figure 3.

The global map of subsurface interfaces identified by the LRS instrument (this work) superposed to the global TiO2 map built from Clementine visible measurements [Lucey et al., 2000] and the shaded relief map from USGS Pigwad server. Identification of one unique subsurface interface (Figures 1b and 2b) is indicated by the red color of the LRS footprints while the identification of two distinct subsurface interfaces (Figures 1c and 2c) is indicated by the yellow color of the LRS footprints. The dashed black contour outlines the lunar maria. With almost no exception, subsurface interfaces are never observed in areas where TiO2 concentration is high. We make the hypothesis that the strong correlation between the LRS detection map and the TiO2 map is due to the strong absorption of low frequency electromagnetic waves by the mineral ilmenite, FeTiO3 (see text for discussion).

Figure 4.

(a) Detections of subsurface features by LRS (red and yellow footprints) superposed to the shaded relief map in the area of Mare Imbrium. (b) Same map of LRS subsurface features superposed to the TiO2 surface concentration map from Lucey et al. [2000]. (c) Same map of LRS subsurface features superposed to the geologic map from Bugiolacchi and Guest [2008]. The dashed black contour outlines Mare Imbrium. As for the global maps, detection of subsurface features in LRS data is systematically associated to low values of TiO2 and thus ilmenite content in the maria lavas. Comparison with the geologic map (c) seems to indicate that areas characterized by the detection of one unique interface (red) or two distinct interfaces (yellow) correspond to two different lower Imbrian series, respectively 2f and 2g, distinguished for the first time by Bugiolacchi and Guest [2008].

[14] The link between the ilmenite distribution on the surface and the detectability of subsurface interfaces at depths of a few hundreds of meters is related to the absorption of the radar wave by ilmenite. This behavior has previously been documented both from laboratory studies of returned Lunar material [Olhoeft and Strangway, 1975] and from radar mapping of the Lunar surface at shorter wavelength [Schaber et al., 1975; Campbell and Hawke, 2005]. An empirical relationship between the concentration of ilmenite in the Lunar regolith and the loss tangent, a measure of the absorption of the electromagnetic wave by media, was derived by Olhoeft and Strangway [1975] on the basis of measurements on Lunar samples returned by the Apollo missions (equation (1)). Energy loss during propagation in an absorbing medium can be calculated using equation (2) from Chyba et al. [1998]. Combining equations (1) and (2), one can calculate the energy loss through propagation in superposed lava flows as a function of the thickness and the concentration in ilmenite of each layer.

equation image
equation image

where tan δ is the loss tangent (dimensionless), ρ is the density (g/cm3), C is the percentage by mass of FeO + TiO2 (ilmenite content), α is the vertical attenuation (dB/m), ɛ is the real part of the dielectric constant (dimensionless) and f is the frequency (in MHz).

[15] The double propagation (down and up) through a 100 meters thick ilmenite-rich (30 wt. % FeTiO3) upper layer results in a total attenuation of 5.8 dB, sufficient to prevent the detection of most of the subsurface features (Figure 2). By comparison, the same value of electromagnetic loss is only obtained after propagation through a 1 km thick layer in the case of an ilmenite-poor (2 wt. % FeTiO3) basalt.

[16] This effect of composition, combined with the consequent scattering that affects the signal (Figure 2), can explain why subsurface interfaces are never observed in the most recent parts of lunar maria as already noted by Oshigami et al. [2009]. The origin of the strong scattering observed in lunar maria is still unclear and subsequent dedicated studies are required to determine if it originates from the surface or the volume of lava layers. As a result, we believe that the absence of detection of shallow and deep subsurface interfaces on large portions of the lunar maria does not indicate that these features are actually absent from these areas. The layering of basaltic flows observed by LRS in ilmenite-poor terrains might well be ubiquitously present within all lunar maria, due to the expected similarity of regolith development as a function of age across the lunar surface. This interpretation is a plausible alternative to the conclusions of Oshigami et al. [2009] who suggest that the variations of the thickness of buried regolith layers cause the heterogeneity in the geographical distribution of subsurface features observed by LRS. Further work is required to evaluate the relative importance of these two possible scenarios.

[17] The presence of regolith layers at relatively constant depth within all lunar maria, as hypothesized in this work, would put interesting constraints on lunar magmatic processes. According to Hiesinger et al. [2002], the average thickness of individual lava flows is 30–60 meters with observed maximal thicknesses of 220 m. Thus, the layering seen by LRS does not correspond to individual lava flows. Only successive lava flows separated by a time lag sufficient to produce a few meters of regolith give rise to the subsurface echoes seen in the LRS dataset. Such a long time lag appears to occur only after a few eruptive events. Comparison with models of volcanism in lunar maria and terrestrial analogs should now be undertaken to understand this behavior and its implications for lunar volcanism.

Acknowledgments

[18] We are grateful to the JAXA team that built and operated the SELENE/Kaguya spacecraft and made this mission possible and successful. We thank the entire LRS team for data calibration and documentation. French national space agency, CNES, provided financial support for this work. We thank Bruce Campbell and an anonymous reviewer for insightful remarks and suggestions that greatly improved this manuscript.

Ancillary