From : C1 Planetary Overview Corrected
2.1 Step-1 Proposal
The Scientific/Technical/Management section of the Step-1 proposal is restricted to one page in length and should include:
The relevance section will be used to confirm that the proposal is submitted to the correct program element.
No evaluation of intrinsic merit will be performed on Step-1 proposals.
The NASA Solicitation and Proposal Integrated Review and Evaluation System (NSPIRES) for proposal submission requires that Step-1 proposals include a summary describing the proposed work; the Scientific/Technical/Management section must be uploaded as the PDF “Proposal Attachment”. Proposers will be notified when they are to submit their Step-2 proposals. If NASA determines that the proposed investigation should instead be submitted to another program, proposers will receive instructions on how to properly submit their Step-2 proposal.
Mixed intro from :
The NASA mission MESSENGER and the ESA mission BepiColombo both carry comprehensive suites of instruments aimed at understanding the surface composition of Mercury. On MESSENGER, these are especially the MASCS and the MDIS instrument - on BepiColombo, the imaging spectrometer MERTIS for the thermal infrared wavelength range, and SYMBIO-SYS for the NIR and VIS ranges from 0.4 to 2 µm. This is be complemented by a suite of x-ray, γ-ray and neutron spectrometer on both spacecraft. The interpretation of the data returned and even the planning of the observations by this suite of instruments will pose a number of challenges. Not only is our knowledge about the surface composition very limited, but we also know little about the texture and the physical properties of the regolith. Therefore it is highly important to use the time until MESSENGER and BepiColombo arrive at Mercury to conduct a thorough scientific preparation with laboratory measurements of emissivity and reflectance spectra of terrestrial and lunar silicates as well as using the opportunity of parallel observations with VIRTIS on VenusExpress during the Venus flyby.
The Dawn mission is designed to investigate the two most massive objects in the main asteroid belt, Vesta and Ceres. These protoplanets are believed to be remnants from the epoch of planet formation, and Dawn’s exploration is intended to help reveal important physical and chemical processes and conditions present at that time and since then.
[…]
Dawn’s scientific measurements at both destinations include panchromatic (in stereo) and multispectral imagery; neutron, visible, infrared, and gamma-ray spectra; and gravity field. To acquire these data, Dawn’s instrument payload comprises a gamma-ray and neutron detector (GRaND), a visible and infrared mapping spectrometer (VIR), and a pair of identical imaging cameras. Gravimetry is accomplished with the telecommunications subsystem, and does not require dedicated flight hardware. Multi-angle images are used to derive topography.
The surface is the interface between the interior and external space. The surface mater- ial is likely to be mostly composed of material from the interior. But the effects of space weathering may have altered the skin of this endogenic material.
The surface may be contaminated by in-falling material not related to or indicative of Ceres and its evolu- tion. However, investigations of the Moon and asteroids suggest that this contamination is not overwhelming. Solar wind-induced space weathering also may have affected the surface material. For example, recently the M3 investigation on Chandrayaan-1 mission reported OH− and perhaps H2O detections in lunar surface material, which could be due to solar wind-induced proton implantation (Pieters et al. 2009;McCordetal. 2011). Evidence of space weathering effect should be sought, both for their own importance and so as to not confuse their effects with endogenic materials and processes.
The VIR will provide constraints on the composition of the surface material, as well as
on its temperature and microstructure (e.g., grain size). These measurements will be espe- cially sensitive to molecular and mineral species. The Gamma Ray-Neutron Spectrometer (GRaND) should also be able to provide constraints on the surface composition. That tech- nique relies on sputtering of material elements from natural γ -ray and neutrons that can then be characterized by a mass spectrometer. This is a more direct approach to determin- ing the composition from the surface as well as from an atmosphere. Combination of the GRaND and VIR data will allow discrimination of the composition from other effects, in particular temperature and microstructure.Depending on its resolution, it may be possible to infer constraints on the thermal state of the icy shell and possible recent tectonic or volcanic events.
target: 200-250 words
current: 278
The integration of the data described above shows that we are forming a more complete understanding of Ceres’ surface composition and can begin to model the processes that led to its formation and evolution.
Those spectral features that appear to be unambiguously identified, interpreted as Mg-carbonates and brucite (Mg(OH)2), are indicative of alteration in the presence of H2Oand CO2.
The strength of their associated spectral features and the amounts in which they are modeled suggest Ceres’ surface, and by extension materials in its interior, has experienced extensive aqueous alteration, possibly more than is recorded in the C1 chondrites.
The very broad near-IR band at ∼1.2 µm that has been attributed to magnetite also supports an interpretation of alteration and partial oxidation of primitive chondritic material.
Several authors, including Li et al.
(2006), Milliken and Rivkin (2009), and Rivkin and Volquardsen (2010) have put forth scenarios in which aqueous alteration oc- curs at rock-water interfaces to form brucite, carbonate, and other alteration minerals.
Such alteration could potentially occur at the base of an ice ocean or at the base of an unmelted ice and silicate crust prior to foundering of that crust.
The presence of Mg-carbonates is consistent with the alteration of brucite in the presence of CO2, where the source of CO2 could be oxidation of CO or graphite originally accreted as such or by the oxidation of carbon from organic material.
The alteration products could then have been brought to the surface by cryovolcanism or after crustal foundering and left behind as a lag deposit after the sublimation of ice, which is unstable at Ceres’ high surface temperatures.
Alternatively, the alteration could have taken place in an original crust which was too porous/low-density to effectively founder.
Geophysical modeling is currently underway to understand which is the more likely scenario.
Whatever process emplaced the alterationminerals atCeres’ surfacewas pervasive.Ceres
shows very little spatial spectral variation in the 0.4–4 µm region relative to other large bod- ies such as Vesta and the Galilean satellites, though spectral variation on smaller asteroids is also rare.
Using ratios of spectral endmember areas on Ceres and the maps of Li et al.
(2006), Rivkin and Volquardsen (2010) concluded that surface features and background re- gions have very similar mineralogies, differing from one another only slightly.
The maps themselves show relatively muted albedo variation, again supporting the idea that Ceres sur- face is by and large a uniform one.
However, it is possible that particle size effects could play a large role on Ceres’ surface.
It is well known that albedo changes can be caused by particle size differences alone, and different wavelength regions will be affected differently, perhaps accounting for the apparent variability seen at mid-IR wavelengths.
However,much additional work still clearly needs to be done.
target: 200-250 words
current: 454
We present observations of Ceres over the 2.2–4.0 lm region taken using the SpeX instrument on the NASA IRTF in 2005. The observations cover Ceres’ entire longitude range and show evidence for a rela- tively uniform surface in terms of Ceres’ composition, however there is a subtle but consistently shal- lower band depth over longitudes associated with bright regions in HST maps, suggesting those areas are slightly less carbonate- and brucite-rich. We also find Ceres’ beaming parameter, a measure of its thermal properties, to have changed with its viewing aspect
In theory, large spectral variation between regions could be
masked on Ceres if that variation were latitudinally symmetrical. However, the maps of surface features found in Li et al. (2006) and Carry et al. (2008a) strongly suggest that is not the case. There- fore, we generally expect that large enough areas of radically dif- ferent spectra would be noticeable in the data analyzed here. While a full exploration of this is beyond the scope of this paper and also necessarily non-unique, we can look at one endmember case based on the strongest absorption band: how small of a bru- cite-free area (following the interpretation that brucite is responsi- ble for the 3.05 lm band) could be detected? If we assume a neutral spectrum for such a region, and assume
that a change of 2% or greater in Ceres’ integrated band depth would be detected (which is comparable to the greatest excursion seen in the real data), that corresponds to a brucite-free area roughly 200 km in diameter at the equator, minimum size increas- ing at higher latitudes as foreshortening becomes important. We can also ask the question in reverse: What does the correla-
tion between the bright area near longitude 130 (Li et al.’s feature number 2, also seen by Carry et al.) and the lower band depth found near those longitudes imply about that surface feature’s true band depth at 3.05 lm? Given the angular size and position of feature 2 as reported in Li
et al. (2006), and including effects due to foreshortening, we esti- mate it contributes over 55% of Ceres’ spectrum when it is at the central meridian seen by an observer. If we ascribe the change in 3.05 lm band depth entirely to this surface feature, it implies a
band depth of ?15–17% compared to the ?20–21% band depth seen on average across the object. This is notable, to be sure, but is not indicative of a wholly different mineralogy. Fig. 6 shows the ratio of the spectrum of the brucite-poorer region to the overall average. This ratio is flat to within a few percent, again suggesting that whatever is responsible for the lesser band depth (whether another mineral or a particle size effect or the like) does not itself have absorptions in this spectral region. Indeed, a mathematical linear mixture of 93% of the spectrum with the deepest brucite band with roughly 7% of a completely flat spectrum results in a close match with the spectrum with the shallowest brucite band.
Observations of Ceres in the 2–4 lm region show subtle but recurring spectral variation with longitude. A higher-albedo region identified in imaging data from HST and ground-based AO from
Keck shows evidence of a shallower band depth (by ?2% of the mean value) at 3.05 lm, identified as due to brucite by Milliken and Rivkin (2009). Variation in the carbonate band near 3.35 lm is also seen, though not at levels greater than 1r from the mean. Albedo variation is not seen, with a scatter roughly equal to the magnitude of any expected variation. Variation in thermal flux is seen between April 2005 and May 2005 and is interpreted with a change in the beaming parameter of Ceres, consistent with the change in phase angle between those observations, though addi- tional variation seen when including other datasets is not consis- tent with phase angle changes. The variation seen in the brucite band depth is roughly equiva-
lent to a variation in 7% in the concentration of a spectrally neutral mineral with the same albedo as Ceres. Ratios of areas with deeper and shallower band depths do not show evidence of additional minerals. If the spectral variation is due to a variation in space weathering, it suggests that the process on Ceres serves to raise albedos rather than lower them.
target: 200-250 words
current: 711
Ceres’ surface is probably one of the most uniform surfaces of small Solar System bodies measured to date, at the ~60km scale of the HST observations.
It didn’t exclude kilometer-scale variation, but it should not be spread on global scale, because it doesn’t affect global albedo.
Nevertheless, this clearly set Ceres apart form known main-belt rocky asteroids: its unique spectrum and a possibly high water content , as indicated by its mean density, point to a distinctive evolution history.
A mix between its distance from the Sun and the high water fraction could have determined an evolutionary path between rocky and icy bodies.
Ceres could have been resurfaced, probably by melted water or ice, after the heavy bombardment phase of Solar System formation.
Ceres is proving to be a very important and unique Solar System object, a key to understanding the early Solar System processes that occurred in the proto-terrestrial planets.
Dawn can measure the mineral composition with visible-IR spectroscopy, and the water-related hydrogen fraction both on the surface and underneath with gamma-ray/neutron spectroscopy. This will help us very much in understanding the history of Ceres.
target: 200-250 words
current: 185
Here disk-resolved observations of Ceres with SINFONI (ESO VLT) in the 1.17–1.32 μm and 1.45–2.35 μm wavelength ranges.
Spatial resolution of ∼75 km on Ceres’ surface.
We do not find any spectral variation above a 3% level, suggesting a homogeneous surface at our spatial resolution.
Slight variations (about 2%) of the spectral slope are detected, geographically correlated with the albedo markings reported from the analysis of the HST and Keck disk-resolved images of Ceres.
The surface of Ceres is thus remarkably homogeneous at our spatial and spectral resolutions
target: 200-250 words
current: 87
We image surface features with diameters in the 50–180 km range and an albedo contrast of ∼6% with respect to the average Ceres albedo.
The spectral behavior of the brightest regions on Ceres is consistent with phyllosilicates and carbonate compounds.
Darker isolated regions could be related to the presence of frost
target: 200-250 words
current: 50
I showed that the long-term evolution of the rocky models introduced by Z09
is determined by the low creep and melting temperatures of some of the hy- drated phases and organics. This leads to rapid compaction of the interior and chemical differentiation accompanied by the segregation of a fraction of ‘‘free’’ water. The end-result model is layered with a rocky core and an outer shell en- riched in hydrated salts, organics, and water ice, a structure similar to that pro- posed by Castillo-Rogez and McCord (2010) starting with different initial conditions. Further modeling of the shell is necessary in order to understand the long-term evolution of the core in this context and quantify the total amount of water released from the silicates. Since the high-density endmember model was developed under the assump-
tion that Ceres could sustain some porosity over the long term, the thermal model presented here invalidates such a model. Devoid with porosity, its density is now at least 10% greater than the upper bound on the observed value. While the low-density chemical model remains consistent with Ceres’ density (upper bound), one must keep in mind that it relies on extreme assumptions on the oxi- dization state of the material accreted in the dwarf planet. The viability of such a scenario and its consistency with available constraints on Ceres’ chemistry, i.e., detection of brucite and carbonates (Milliken and Rivkin, 2009) remain to be tested.
This study does not question the possibility that smaller C-class asteroids may
be made up of porous carbonaceous chondrite material though. It simply points out that such an assemblage is likely to be unstable in an object of Ceres’ size, which could achieve moderate to high temperatures in the course of its evolution. A model in which water is partly in the form of ice appears most consistent with the dwarf planet’s properties, which puts Ceres in the same class as the Themis parent body (Castillo-Rogez and Schmidt, 2010). More generally, this work is also relevant to other studies that have suggested a
chondritic composition for the cores of large icy satellites (e.g., Fortes, 2011, for Titan). The destabilization of a large fraction of the hydrated material is warranted at the temperatures expected in these large cores, with potentially significant ther- mal and mechanical consequences
target: 200-250 words
current: 370
To reveal hidden structure in the unlabeled or unsupervised data, we applied a hierarchical clustering analysis technique. Cluster analysis, or clustering, is the task of grouping a set of objects into clusters according to shared characteristics or properties [Everitt, 1993; Michaud, 1997].
To characterize the spectral units and to expose geographical correlations among units, we used only the spectral subset of six reflectance bands (Figure 6), yielding a vectorial space with a lower dimensionality than that of the original spectral data set. We then estimated the pairwise distance between two spectra S1 and S2 by a Chebyshev distance metric s defined as max (|S1,i – S2,i|) where i = 1,…6 denotes the reflectance band used. This distance was used to compute the hierarchical clustering of the multi-dimensional data points by a weighted centroid approach [de Hoon et al., 2004]. In this method, the distance between two clusters is defined as the distance between the centroids of each cluster, and the centroid of a cluster is defined by the average position of all the sub-clusters, weighted by the number of objects in each group.
The clustering technique applied here allows for the identification of units on the surface that show similar spectral characteristics in the chosen bands. The maximum distance between neighbors within a given cluster defines the “similarity” of data in that cluster. Unlike other clustering methodologies, we did not set the number of partitions a priori (meaning we did not assume a given number of distinct spectral units), nor did we search for the optimal partitioning of the data set (at least at this stage), because these approaches could have been too sensitive to errors not yet corrected in the data. Instead we cycled through all possible partitions of the dataset, which are defined independently of the dataset structure.
We defined a “classification tree” as the collection of all the partitions with boundaries given by the partition where each data point constitutes a cluster on its own and the partition where there is one cluster including all the data point (in which case the cluster centroid coincides with the data mean). We could then observe the evolution of each single cluster stepping (with decreasing Chebyshev distance) in the cluster tree and examine how the clusters were interrelated and when they broke into smaller groups. This assessment can be conducted on a geographical basis, to examine the correlation of spectral cluster locations with imaged features or regions defined on the basis of other quantities or measurements. Locations of a given cluster, for instance, might correlate with incidence, emission, or phase angle, the angle between the instrument boresight and the Sun–spacecraft vector, spacecraft altitude, or sensor temperature. We found no statistically significant correlation in our analysis, however, between cluster locations and phase angle, incidence angle, emission angle, or VIS detector temperature.
target: 200-250 words
current: 467
Machine learning - Wikipedia, the free encyclopedia
Machine learning is a subfield of computer science (CS) and artificial intelligence (AI) that deals with the construction and study of systems that can learn from data, rather than follow only explicitly programmed instructions. Besides CS and AI, it has strong ties to statistics and optimization, which deliver both methods and theory to the field. Machine learning is employed in a range of computing tasks where designing and programming explicit rule-based algorithms is infeasible for a variety of reasons. Example applications include spam filtering, optical character recognition (OCR),[1] search engines and computer vision. Machine learning is sometimes conflated with, and sometimes distinguished from data mining and pattern recognition.[citation needed]
Machine learning tasks can be of several forms. In supervised learning, the computer is presented with example inputs and their desired outputs, given by a “teacher”, and the goal is to learn a general rule that maps inputs to outputs. Spam filtering is an example of supervised learning, in particular classification, where the learning algorithm is presented with email (or other) messages labeled beforehand as “spam” or “not spam”, to produce a computer program that labels unseen messages as either spam or not.
In unsupervised learning, no labels are given to the learning algorithm, leaving it on its own to groups of similar inputs (clustering), density estimates or projections of high-dimensional data that can be visualised effectively.[2]:3 Unsupervised learning can be a goal in itself (discovering hidden patterns in data) or a means towards an end. Topic modeling is an example of unsupervised learning, where a program is given a list of human language documents and is tasked to find out which documents cover similar topics.
In reinforcement learning, a computer program interacts with a dynamic environment in which it must perform a certain goal (such as driving a vehicle), without a teacher explicitly telling it whether it has come close to its goal or not.
target: 200-250 words
current: 316
target: 200-250 words
current: 156
The NASA mission DAWN carry a suites of instruments aimed at understanding the two most massive objects in the main asteroid belt: Vesta and Ceres. These protoplanets are believed to be remnants from the epoch of planet formation, and Dawn’s exploration is intended to help reveal important physical and chemical processes and conditions present at that time and since then. DAWN has already successfully completed the exploration of Vesta September 2012 and it is now on the journey to Ceres. The interpretation of the data returned and even the planning of the observations by this suite of instruments will pose a number of challenges. Not only is our knowledge about the surface composition very limited, but we also know little about the texture and the physical properties of the regolith. Therefore is essential to plan analysis strategies that could exploit the hidden relationship buried in the upcoming data and explore relationship among all the complementary dataset.
target: 200-250 words
current: 260
The surface is likely to be mostly composed of material from the interior, likely altered by the effect of space weathering. Moreover it could be contaminated by in-falling material unrelated to Ceres and its evolution. Space weathering effect should be taken in account, both for their inherent importance and also to not confuse with local processes.
Among the few know key point of Ceres, :
To the date, some spectral features are unambiguously identified, and interpreted as Mg-carbonates and brucite, of alteration in the presence of H2O and CO2.
Ceres shows very little spatial spectral variation in the 0.4–4 µm region
Using ratios of spectral endmember areas on Ceres the surface features and background regions have very similar mineralogies, with very subtle differences.
Also albedo maps show dampen variation supporting the idea that Ceres surface is by and large a uniform one, even if particle size effects could play a large role.
In this frame DAWN will explore Ceres with the visible and infrared mapping spectrometer (VIR), the will constraints the composition of the surface, the temperature and linked morphological parameter like the regolith grain size. The gamma-ray and neutron detector (GRaND) will provide information on the surface composition and the Framing Camera (FC) will attempt to determine the origin and evolution of Vesta and Ceres by mapping the extent of geologic processes on the asteroid surfaces, and by using the cratering record. Depending on final measurement resolution, it may be possible to infer constraints on the thermal state of the icy shell and possible recent tectonic or volcanic events.
target: 200-250 words
current: 786
The small spectral variation resemble the problem posed by the spectral data returned by the MESSENGER mission. The three year mission around Mercury return essentially a spectral scenario hard to decipher, and the still ongoing analysis on the phase function force us to develop methodologies that could exploit the hidden structure in the data despite instrumental effect, lacking phase function, strong solar line effect, very low reflectance and substantially attenuated surface variation. The initial theories of an essentially Moon like body were drastically set aside, and today it is not certain which surface material contribute to the extremely dark and spectrally flat surface of Mercury.
To tackle this problem and decipher the data, conventional methods would compare the data to known sample (e.g. direct search for match in existing spectral database) of would try fit the make with model that could be calculated from known endmember (e.g. from simple linear mixing model to radiative transfer equation with different scattering models through solid media).On Mercury , these methodologies were set aside because inapplicable or incapable of produce meaningful results.
To reveal hidden structure in the data we chose to access the large family machine learning methods. Machine learning is a subfield of computer science and artificial intelligence that deals with the construction and study of systems that can learn from data, rather than follow only explicitly programmed instructions. It has strong ties to statistics and optimization, which deliver both methods and theory to the field.
Machine learning tasks can be of several forms: supervised learning, were the machine is “teached” to learn how to construct general classification rules starting from a small tranining dataset with know labels attached. In unsupervised learning, no labels are given to the learning algorithm, leaving it on its own to groups of similar inputs (clustering), density estimates or projections of high-dimensional data that can be visualised effectively. Unsupervised learning can be a goal in itself (discovering hidden patterns in data) or a means towards an end. In reinforcement learning, a computer program interacts with a dynamic environment in which it must perform a certain goal, without a teacher explicitly telling it whether it has come close to its goal or not. The particular problem characteristic led to develop an unsupervised classification method based on the hierarchical clustering of the data. Cluster analysis, or clustering, is the task of grouping a set of objects into clusters according to shared characteristics or properties, in this case spectral, chemical or visual character of the surface. Unlike other clustering methodologies, we did not set the number of partitions a priori (meaning we did not assume a given number of distinct spectral units), nor did we search for the optimal partitioning of the data set, because these approaches could have been too sensitive to errors not yet corrected in the data. We developed a mixed exploratory-clustering approach, where we cycled through all possible partitions of the “classification tree” of the dataset. We could then observe the evolution of each single cluster stepping in the cluster tree and examine how the clusters were interrelated and when they broke into smaller groups. This assessment can be conducted on a geographical basis, to examine the correlation of spectral cluster locations with imaged features or regions defined on the basis of other quantities or measurements.
For Mercury we were able to show the existence of homogeneous spectral global megaregion, sharing commons spectralproperities. Comparison with the MESSENGER camera confirm this global characteristics of Mercury.
This technique has the advantage to be able to treat as input data a fusion of different dataset. On Mercury, we successfully merged spectral and chemical data to simultaneously characterize different aspects of the surface .
Our approach could successfully identify spectro-chemical region on the surface, that lead to the discovery of unseen spatially coherent region at global and local scale. As example, our global classification maps identify as spurious point areas recognised as pixel bering hollow structure, typical of Mercury,
The scenario is similar in many aspect to the expected data from Ceres, thus our approach could empower the knowledge returned form Ceres’ data in a drastically different way than conventional methods.
We can exploit the hidden relationship buried data, potentially invisible to more traditional approach, and explore relationship among all the complementary dataset, exploiting the ful potential of the DAWN data.
We plan thus to use VIR, GRand and Framing Camera to define spectral, chemical and visual region could, resampling the dataset to an homogeneous resolution on the surface and fused in layers representing the spectral, chemical and visual characteristic of each point on the surface.
This could potentially bring to discover data structure with spatial coherence that could be remain unexplored by any other method tackling a single dataset.