Spectral Clustering on Mercury Hollows: the Dominici Crater Case
Total Page：16
File Type：pdf, Size：1020Kb
Lunar and Planetary Science XLVIII (2017) 1329.pdf SPECTRAL CLUSTERING ON MERCURY HOLLOWS: THE DOMINICI CRATER CASE. A. Lucchetti1, M. Pajola2,3,1, G. Cremonese1, C. Carli4, G. A. Marzo5 and T. Roush3, 1INAF-Astronomical Observatory of Padova, Vicolo dell’Osservatorio 5, 35131 Padova, Italy ([email protected] ), 2Universities Space Research Association, NASA NPP Program (Supported by an appointment at NASA Ames Research Center: [email protected]), 3NASA Ames Research Center, Moffett Field, CA 94035, USA; 4INAF-IAPS Roma, Istituto di Astrofisica e Planetologia Spaziali di Roma, Via del Fosso del Cavaliere, 00133 Rome, Italy; 5ENEA Centro Ricerche Casaccia, 00123 Rome, Italy. Introduction: The Mercury Dual Imaging System [8], i.e. incidence angle of 30°, emission angle of 0° (MDIS, [1]) onboard NASA MESSENGER (MErcury and phase angle of 30°. On the photometrically cor- Surface, Space ENvironment, GEochemistry, and rected dataset we applied a statistical clustering over Ranging) spacecraft, provided the first global coverage the entire dataset based on a K-means partitioning al- of Mercury's surface with varying spatial resolution. gorithm [9]. It was developed and evaluated by [9-11] Early in the mission, high-resolution images showed and makes use of the Calinski and Harabasz criterion that specific areas exhibiting high reflectance and rela- [12] to find the intrinsically natural number of clusters, tive bluer in color were composed of shallow, irregular making the process unsupervised. A natural number of and rimless, flat-floored depressions with bright interi- ten clusters was identified within the crater and its ors and halos, often found on crater walls, rims, floors closest surroundings, see Fig. 2. Each resulting cluster and central peaks [2,3,4]. These features were named is characterized by an average multi-color spectrum, “hollows”: they are fresh in appearance and may be and its associated variability. actively forming today via a mechanism that involves depletion of subsurface volatiles [2,5]. Understanding the composition of these features provides additional information on Mercury’s surface characterization. For this purpose, we applied a spectral clustering method to the ~20 km wide Dominici crater [6], Fig. 1. We chose this target due to previous different spectral detection [6]. Fig 1. A: the WAC image (EW0210848973D) showing the Dominici crater (center latitude = 1.38°N, center longitude = 323.5°E). B: the NAC image (EN0253965560M) showing hollows both in the center and in the south crater wall/rim of the Dominici crater (white arrows). Dataset and Methodology: The MDIS imager was Fig 2. A: the WAC reference image (EW0210848973D). B: equipped with a monochromatic narrow angle camera The 10 clusters identified on the Dominici MDIS dataset. (NAC), and a multiband wide angle camera (WAC), used to investigate the surface composition. Here we This approach has been previously applied for used the WAC dataset covering Dominici crater with a compositional interpretation of different Solar System scale of 935 m/pixel through eight filters, ranging from objects, e.g. asteroids, Mars, Phobos and Iapetus [11, 0.433 to 0.996 μm. 13, 14, 15]. The algorithm is agnostic of the physical The images have been photometrically corrected with meaning of the resulting clusters, and scientific inter- Hapke methods [7], using the parameters derived in pretation is required for their subsequent evaluation. Lunar and Planetary Science XLVIII (2017) 1329.pdf Fig 3. The average normalized spectra derived on all the identified clusters. Colors refers to Fig. 2. The legenda shows the cluster # and its not normalized I/F value at 0.558 μm. Results: As it is possible to see from Fig. 2 the maybe a mixture between the previously mentionated clustering technique spectrally separates the Dominici terrains (MgS, [6]), or new compositional units. surrounding terrains (Clusters 0, 1, 2, 6), from its inte- rior (Clusters 7, 8), as well as the two hollows (Clusters Acknowledgments: This activity has been realized 3, 4, 5) from their edges (Cluster 9). The resulting under the BepiColombo ASI-INAF contract mean spectra for each derived cluster are presented in no.I/022/10/0. M. Pajola was supported for this re- Fig. 3. search by an appointment to the National Aeronautics Cluster 3 and 5 show the most peculiar spectra. and Space Administration (NASA) Post-doctoral Pro- These spectral units are located on the brightest part of gram at the Ames Research Center administered by the south wall/rim of Dominici crater, and clearly pre- Universities Space Research Association (USRA) sent a wide absorption band between 0.558 and 0.828 through a contract with NASA. μm, and a hint of absorption towards the IR. This de- tection is similar to what was described in [6], even if it References: [1] Hawkins, S. E. et al. (2007), Space is not located in the crater center as previously report- Sci. Rev., 131, 247-338 [2] Blewett D. T, et al. (2011) ed, that here is spectrally mapped by cluster 4. Science, 333, 1856–1859. [3] Blewett D. T. et al. Clusters 0, 1, 2, 6 typically present a red slope in the (2013) JGR Planets, 118, 1013-1032. [4] Thomas R. J. VNIR with a possible weak absorption band centered et al. (2014), Icarus, 229, 221–235. [5] Vaughan W. at 0.748 μm. Clusters 4, 7, 8 and 9 show almost no M. et al. (2012) LPSC, 43, abstract 1187. [6] Vilas, F. absorption between 0.558 and 0.828 μm, but a possible et al. (2016), GRL, 43, 1450-1456. [7] Hapke, B. absorption towards the IR is still evident. Clusters 4, 7, (2002), Icarus, 157, 523-534. [8] Domingue, D. et al. 8 and 9 seem to be representative of a possible “spec- (2015), Icarus, 257, 477-488. [9] Marzo, G. et al. tral transition” between the two groups. (2006) JGR, 111, E03002. [10] Marzo, G. et al. The application of the clustering technique permits (2008), JGR, 113, E12009. [11] Marzo, G. et al. to study in deeper detail the spectral differences pre- (2009), JGR, 114, E08001. [12] Calinski, T., sented in [6]. In particular by means of this method we Harabasz, J., (1974), Commun. Statist. 3, 1–27. [13] discriminated areas with a possible diagnostic absorp- Pinilla-Alonso, N. et al. (2011), Icarus, 215, 1, 75. [14] tion indicative of sulfides (e.g. MgS as suggested by Dalle Ore, C. et al. (2012), Icarus, 221, 2, 735. [15] [6]). Moreover we were able to separate the surround- Pajola, M. and Roush, T. (2016), American Astronom- ing LRM unit (low reflectance material), but also pos- ical Society, DPS meeting #48, id.428.05. sibile intermediate terrains with different spectral prop- erties which could be “spectral transition” terrains, .