5.1. Mare Crater Spectra and Implications for Basalt Mineralogy
 The reflectance properties of both mare soils and fresh craters observed by M3 confirm the presence of a strong 1 μm absorption and weak 2 μm band within the western high-Ti basalts [Pieters et al., 1980; Staid and Pieters, 2001; Lucey, 2004]. The M3 data, however, provide a vastly improved combination of spectral and spatial resolution to directly observe the reflectance properties of small craters excavating individual compositional units within the maria. These observations allow for improved discrimination of potential components which share similarly positioned absorptions, such as olivine and iron-rich glass, and examination of specific stratigraphic sequences of basalts and their mineralogic associations.
 Olivine reflectance spectra display a broad and asymmetric composite absorption near 1 μm and lack a 2 μm absorption that is present in pyroxenes (Figure 4c) [Adams, 1975; Singer, 1980; Burns, 1993]. A central absorption just beyond 1 μm is caused by iron in the M2 site, and ‘wing’ absorptions near 0.9 and 1.3 μm result from iron in the M1 site [Burns, 1993]. In mineral mixtures with olivine, an asymmetric band near 1 μm, with a ‘secondary’ band near 1.3 μm, may be most visible [Singer, 1981; Pieters et al., 1980]. Within the global mode data examined, area 2 (youngest high-Ti unit north of Aristarchus in Figures 2 and 3) displays the broadest 1 μm absorption and weakest 2 versus 1 μm band strengths among the mare basalts. The 1 μm absorption associated with these basalts is much broader than in the neighboring pyroxene-rich basalts of area 1 and is centered beyond 1 μm. A longer-wavelength absorption centered between 1.2 and 1.3 μm is also apparent in this spectrum. The high-Ti basalts that include area 2 are, therefore, interpreted to contain the most visible and presumably highest component of olivine of the three mare regions studied in the global mode data. Another region with a strong 1 μm band but a weak 2 μm absorption in Figure 3 occurs in southeastern Aristarchus, a region previously interpreted as containing olivine [Le Mouélic et al., 1999]. This area is examined in more detail using M3 data in the companion paper by Mustard et al. .
 Other common lunar minerals that produce absorptions beyond 1 μm and lack significant 2 μm features include plagioclase feldspar and Fe-rich glass. Plagioclase feldspar has weak absorptions at 1.2 μm and lunar glasses that have a broad absorption centered around 1 μm [Adams, 1975]. However, weak plagioclase absorptions are not expected to be visible in the presence of even small amounts of darker mafic minerals present in basalts [e.g., Crown and Pieters, 1987] and lose absorption features near 1 μm when shocked, such that the mineral has proven difficult to detect even in most highland regions. Fe-rich lunar glasses also contain broad absorptions extending beyond 1 μm; however, these absorptions are more symmetrical than the composite absorption observed within olivine. Fe-rich glasses produced by rapid cooling of basalts could contribute to broad absorptions observed in the weak spectral features associated with the western high-Ti soils [Pieters et al., 1980]. However, FeO-rich glasses are less likely to produce a strong and distinctly olivine-like signature in fresh craters excavating and mixing materials from depth, as observed in the M3 data of area 2.
 Olivine is less absorbing than pyroxenes and is thus likely to be masked by pyroxene absorptions, unless the olivine is present in relatively large abundance [Singer, 1981; Pieters et al., 1980; Mustard and Pieters, 1987]. The presence of the distinct olivine shape within spectra from area 2 suggests that either the olivine is very abundant relative to pyroxene (olivine/pyroxene > 1) or factors such as grain size and mineral associations within these basalts allow light to reflect more easily from the olivine-rich component. Since darkening components such as ilmenite may be associated with the pyroxene component or matrix of a basalt, independently of crystals of olivine, these basalts may have a lower olivine/pyroxene ratio than would be inferred by a linear interpretation of the strength of the olivine features observed in the M3 data.
 The high-Ti basalts sampled in area 3 lie west of Aristarchus and appear as a light red hue in the M3 IBD mosaic in Figure 2. The distribution of these basalts in both Procellarum and Mare Imbrium, as well as comparisons to mare age estimates, indicates that they generally predate the spectral unit sampled at area 2. These mare soils also have relatively weak 2 μm absorptions compared with older, surrounding low-Ti basalts, but the differences in band strength are less extreme than observed for the basalts sampled in area 2. The spectral properties of fresh mare craters sampled from area 3 (Figure 4), also display a broad, long-wavelength 1 μm band and weaker 2 μm absorptions consistent with the presence of some olivine. The spectral properties of the ferrous bands within the area 3 basalts appear intermediate between those of the pyroxene-rich, low-Ti basalts and the youngest high-Ti basalts. Area 3 high-Ti basalts are, therefore, interpreted to have at least some olivine present, but a lower average olivine/pyroxene ratio than the basalts sampled in area 2. Alternatively, differences in grain size and associations between olivine and the opaques (e.g., chromite, ilmenite) within these basalts could also result in the observed spectral differences from area 2 with similar olivine contents. Exposure to greater amounts of vertical mixing with the underlying and pyroxene-rich low-Ti basalts could also result in lower inferred olivine contents than the stratigraphically younger high-Ti basalts. However, the global data sampled at area 3 have a similar spectrally blue slope like area 2 and do not display an increase in albedo or reddening that would be expected from such mixing of mare materials.
 The targeted M3 data near Lichtenberg crater provides an opportunity to examine the shape of the 1 μm feature within the youngest high-Ti basalts in greater detail. Unlike the global mode spectra presented in Figure 4, these data were thermally corrected using the approach of Clark et al. , which should provide improvements in the shape of the data near 2 μm, relative to the uncorrected data. These basalts are spectrally contiguous with area 2 in the global IBD parameter mosaic, but as described previously, are interpreted to be very thin in the region near Lichtenberg. The comparison in Figure 6 attempts to isolate some of the most olivine-rich basalts associated with craters in the high-Ti unit for comparison to the older, low-Ti and pyroxene-rich basalts to the west. It is again noted that the calibrations of the targeted data used for this study are very preliminary and a local smoothing of the data based on the EFFORT method was necessary to make these preliminary comparisons using the targeted data. The phase angles of the data themselves were quite high (∼60°), also making comparisons of samplings of crater materials complicated by shadowed and illuminated slopes. The resulting spectra for both the high- and low-Ti units, however, are consistent with craters sampled in the lower spatial and spectral resolution global data. In particular, the low-Ti basalts in this region have spectral properties that are consistent with laboratory and telescopic measurements of typical pyroxene-rich lunar basalts. The crater materials sampled from the high-Ti unit have spectral properties that are consistent with a high abundance of olivine. Due to the limited locations of fresh crater ejecta that could be sampled in this small region of targeted data, it is not possible to determine how typical these olivine rich materials are within other craters sampling this spectral unit.
 Previous studies of Lunar Prospector and Clementine data have characterized the late stage western, high-Ti basalts as among the most FeO-rich basalts on the Moon [Lawrence et al., 2002; Staid and Pieters, 2001]. Determining whether the olivine compositions of these basalts are also relatively FeO-rich is relevant to the evolution and source regions of these basalts. The shape and position of the 1 μm composite absorption for olivine are known to vary systematically from MgO- to FeO-rich compositions due to the position of Fe2+ in the crystal structure [Burns, 1993; Sunshine and Pieters, 1998]. The systematic changes in olivine band positions and shape with composition are well documented, and detailed modeling of olivine has been demonstrated in the laboratory and in remote sensing data [Burns, 1993; Sunshine and Pieters, 1998; Sunshine et al., 2007]. Qualitative comparisons of the spectra in Figure 6 to laboratory spectra of MgO-rich and FeO-rich (fosterite and fayalite) olivines indicate that the late stage, high-Ti basalts may be relatively FeO-rich. In particular, the longer-wavelength M2 absorption appears comparatively strong in these basalts relative to their central M2 absorption, producing an overall 1 μm band shape and long-wavelength edge more similar to the FeO-rich than MgO-rich end-members measured in laboratory studies [Sunshine and Pieters, 1998]. However, the modeling of absorption band positions in lunar basalts containing both olivine and pyroxene is a complex problem and beyond the current scope of this paper. Furthermore, grain size can also affect the shape of olivine spectra and complicate the identification of MgO-rich versus FeO-rich olivines [e.g., Lucey, 1998]. As a result, no conclusions about the composition of the olivine in these basalts can be reached without additional calibration and modeling of the M3 targeted data. The presence of high-FeO basalts can be expected to have lower Fo contents than normal but probably not lower than Fo50 in quantities detectable with remote sensing data.
5.2. Stratigraphy and Distribution of Olivine-Rich Basalts
 The inferred abundance of olivine in the western high-titanium basalts is observed to vary stratigraphically, with the uppermost flows in several areas (dashed lines, Figure 2) displaying the broadest 1 μm absorptions and weakest relative 2 μm band strengths. Stratigraphically older high-Ti flows within Imbrium and Procellarum (e.g., area 3 in Figure 2) also appear to contain olivine but at lower or more variable concentrations relative to their pyroxene abundances. This stratigraphic pattern is observed in a number of different regions with a wide range of estimated ages [Hiesinger et al., 2003, 2011]. Some of the youngest dated basalts near Lichtenberg crater [Schultz and Spudis, 1983; Hiesinger et al., 2003, 2011] are included in and contiguous with the most olivine-rich group, sampled in area 2 in global mode data and near Lichtenberg in the M3 targeted data. Additionally, the uppermost flows of basalts in some regions with older age estimates, such as in central Mare Imbrium and areas of southern Procellarum, also have similar spectral properties to these olivine-rich basalts. For example, the distribution of flows with increasing inferred olivine content are also a good match to phases of basalt emplacement mapped by Schaber et al.  and support previous interpretations of increasing olivine abundance with subsequent eruptive phases of iron and titanium-rich basalts in this region [Staid and Pieters, 2001].
 The volcanic history of related basaltic units in the Marius Hills complex (bottom of Figure 2) is examined in more detail in a companion study of M3 data by Besse et al. . This study also identifies olivine-rich basalts within the youngest regional high-Ti flows in the crater Marius and surrounding deposits. Together, these new M3 observations suggest that the late stage basalts exhibit a pattern of increasing olivine abundance with subsequent emplacement that is widespread within the western high-Ti deposits, producing recognizable sequences of mare volcanism over more than a billion years of lunar history. Future studies of sequential phases of these basalts, such as those mapped by Schaber [1973a] in Mare Imbrium represent important sites for trying to substantiate and expand on these observations. For example, the apparent stratigraphic evolution and Fe-rich compositions of these basalts suggest an origin through evolved residual melts rather than through the assimilation of more primitive (Mg-rich) olivine-rich sources. If future studies confirm that these basalts have moderate to high-FeO contents that increase with subsequent eruptive phases, such basalts might originate from a mantle magma chamber that is undergoing mineral fractionation and crystal settling, resulting in increased FeO contents to the residual magma, which might be tapped at different times. This would result in increased FeO contents for subsequent basaltic olivines, and with concurrent decreases in abundances. Alternatively, if olivine FeO abundances are found to decrease with subsequent eruption, this may indicate that source regions are deepening over time or that a single olivine-rich source is being melted repeatedly.
 The late stage western basalts in Imbrium and Procellarum (which extend into northwestern Frigoris) represent the largest exposure of this unique spectral type that we have seen on the Moon. Preliminary examination of global-scale M3 parameter images does not reveal similar spectral properties among the other large-scale mare deposits on the Moon. However, small deposits with similar spectral characteristics can be identified in M3 parameter images within several areas outside of Mare Imbrium and Procellarum. Where present, these deposits appear to occur as small flows near or superimposed on older and more extensive maria or as isolated mare ponds. One such region occurs within small mare ponds southwest of Humorum where deposits that appear as nondescript mare ponds in Clementine color ratio images display strong 1 μm and weak 2 μm bands in M3 IBD images. Other small deposits that stand out in the M3 IBD parameter occur along the edges of Mare Frigoris. However, because the global M3 parameter images are nonunique, each potential basalt occurrence will require further examination using the full spatial and spectral resolution M3 data before they can be confirmed to represent basalts with similar spectral properties as those observed in this study.
 M3 observations of western nearside maria have provided new information about the mineralogy and emplacement history of these last major phases of lunar volcanism. The M3 data have confirmed previous observations that these basalts exhibit a unique combination of strong 1 μm and weak 2 μm ferrous bands [Pieters et al., 1980; Staid and Pieters, 2001; Lucey, 2004] and have observed these properties throughout large regions of soils and craters within these deposits. The improved spatial and spectral resolution of the M3 data allows direct observation of small craters sampling individual basaltic flows and improved discrimination of diagnostic absorption bands for the interpretation of mineralogy. Based on these new observations, we interpret the western basalts as having significant and variable quantities of olivine that can be directly observed in the reflectance properties of fresh mare crater regoliths. The abundance of olivine within these basalts appears to vary stratigraphically, with the uppermost flows being most olivine rich. This mineralogical trend appears to exist across regions and absolute ages of emplacement suggesting a common pattern of magma evolution of within these high-Ti basalts. Some small mare deposits with similar spectral properties may exist as the final products of mare volcanism in other isolated regions of the Moon; however, large regions of basalts with similar compositions (high titanium, high iron, and abundant olivine) are not observed as other major mare deposits elsewhere on the Moon.
 This initial study of the M3 data for basalts on the western nearside only begins to explore the wealth of information provided by these new imaging spectrometer measurements. The discrete stratigraphic sequences of the high-Ti basalts occurring in Mare Imbrium and portions of Procellarum will be important targets for future study. More detailed examination of M3 data within such previously mapped flows should provide information about mineralogical trends related to the evolution of mare basalt composition between eruptive phases as well as potential changes due to fractionation of materials prior to and during emplacement. Several concurrent studies of M3 data have demonstrated the utility of quantitative techniques such as the Modified Gaussian Method (MGM) to interpret olivine and pyroxene compositions from M3 data [e.g., Isaacson et al., 2011; Klima et al., 2011]. Future work will attempt to provide estimates of the relative abundance and composition of pyroxene and olivine to provide further constraints on their source regions, temporal evolution and emplacement mechanisms.