VIEWS FROM EPOXI: COLORS IN OUR SOLAR SYSTEM AS AN ANALOG FOR EXTRASOLAR PLANETS

One of the earliest studies of Earth for the purpose of understanding whether its habitability could be remotely determined was undertaken by the Galileo spacecraft as it flew by Earth during a gravity assist maneuver on the way to its primary mission at Jupiter (Sagan et al. 1993). The data collected on this flyby included spatially resolved UV and NIR spectra and visible photometry, as well as observations at radio wavelengths. Whole-disk high-resolution spectroscopy at visible wavelengths has since been acquired by the Visible and Infrared Thermal Imaging Spectrometer (VIRTIS) aboard Venus Express, but has yet to be analyzed (Coradini et al. 1998; Grinspoon et al. 2008).

Photometric studies of the Moon have until recently been restricted to ground-based observations and thus limited to only the lunar near side. McCord et al. (1972), using low-resolution filter photometry, studied the spectral characteristics of various near side lunar regions including mare, upland, and bright craters using ground-based telescopes calibrated relative to laboratory-reflected light measurements of Apollo 11 returned samples. Observations of the lunar surface were collected with the 21 inch and 60 inch telescopes at the Mount Wilson Observatory using 24 medium-band filters, between 0.3 and 1.1 μm. The McCord data are characterized by variations in the lunar surface reflectance between 800 and 1000 nm. See Section 4 for more details.

Three spacecraft missions observed the lunar far side through narrow- to medium-band filters similar to those of the High Resolution Instrument (HRI) used in the data discussed herein. The data from Galileo's solid state imaging cameras (Belton et al. 1992), Clementine's UV/Visible instrument (Hillier et al. 1999), and SMART-1 orbiter's visible and NIR camera AMIE (Advanced Moon micro-Imager Experiment; Josset 2008) were all at higher spatial resolution than the EPOCh images though no single data set spanned the UV to NIR spectral regions as completely as the HRI.

Multiple full-disk brightness measurements of Mars have been collected with ground-based telescopes (e.g., Younkin 1966; Irvine et al. 1968a, 1968b; McCord & Westphal 1971), but they are limited by atmospheric interference and do not cover a complete rotation of Mars. Visible spectroscopic observations of Mars have been collected by Mars Global Surveyor and are not contaminated by terrestrial atmospheric effects (e.g., Rogers et al. 2007), but they are at higher spatial resolution than the EPOCh Mars data and are not full-disk observations. The EPOCh data set therefore afforded the unique opportunity to observe the full disk of Mars over a full revolution without the effects of neither Earth's nor Mars' atmosphere.

The EPOCh investigation planned five observations of Earth for 2008. EarthObs4, presented here, was conducted over 24 hr beginning approximately on 20:00 UT 2008 May 28 and included a transit of Earth by the Moon; see Figure 1(a). This sequence provided contemporaneous photometry and spectroscopy of the Moon and Earth. It also allowed for observations of the far side of the Moon, which is not visible from Earth due to the Moon's synchronous rotation. The illuminated region of the Moon in the 2008 May images spanned from 160° W to 260° W longitudes (approximately half of the lunar far side) and the phase angle for both Earth and Moon averaged 751. An additional set of calibration data from 2005 January 16 was used to analyze the near side lunar reflectance for comparison with the far side data. Approximately 2/3 of the illuminated surface of the front side of the Moon was visible in the calibration images and encompassed the southeast near side Balmer crater regions; see Figure 1(b) (Maxwell & Andre 1981). The phase angle for these images was 976. In addition to lunar and Earth data, we analyzed 64 sets of full-disk observations of Mars collected over a 24 hr period starting on 2009 November 20. The observations covered a nearly complete rotation period of Mars at a phase angle of 37° (Figure 1(c)). Figure 2 shows the individual data sets for each body and Table 1 contains details on observation parameters for all data sets. The variations in the wavelength-dependent brightness of the Moon from the different observations shown in Figure 2 are relatively small. For Mars, the variations are larger and are due to the planet's rotation causing different surfaces with distinct compositions rotating into and out of view. Similarly, for Earth the variations are due to the changing scene including cloud coverage and structures, continents, and oceans on the illuminated disk as Earth rotates. The variability demonstrated in Figure 2 does not affect the relative reflectance values used in this work to characterize the worlds.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The EPOCh observations were taken using the HRI on the DI flyby spacecraft (Hampton et al. 2005), a 0.3 m f/35 optical Cassegrain telescope equipped with a 9-position filter wheel and a CCD camera. Two of the positions have broadband white-light filters centered at 650 nm and the other seven are medium-band filters with 100 nm bandwidths evenly spaced between 300 and 1000 nm (Hampton et al. 2005). Images of Earth and the Moon were taken through the 350, 750, and 950 nm filters hourly and through the remaining medium-band filters every 15 minutes, with exposure times ranging from 8 to 73 ms depending on the filter throughputs. See Table 2 for a list of filter center wavelengths and exposure times. We used 8 of the 12 sets of lunar transit observations for our analysis. The remaining four sets were discarded because either the Moon was transiting Earth or the gradient of scattered Earthlight across the Moon was too large for a single background value subtraction.

The reflectance of the lunar far and southeast region of the near side, shown in Figure 3, increases with a relatively linear slope from the UV to the NIR except in the 950 nm filter of the far side. In this bandpass, the disk-integrated far side exhibits an absorption feature not present in the summed images of the near side region observed by the DI spacecraft. The variation in the 950 nm filter is consistent with narrowband photometry of lunar regions from McCord et al. (1972) presented in Figure 4. Absorptions are seen in the McCord data for the lunar standard region Mare Serenitatis 2 (d) and are also prominent in other bright crater regions (e, f, and g). The disk-integrated reflectance of Mars shown in Figure 2 is also characterized by a nonlinear increase from the UV to the NIR with an absorption feature at 950 nm.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The Moon and Mercury are airless bodies and Mars' atmosphere is transparent in the visible and NIR wavelengths. Consequently, their reflected light is dominated by the surface composition and the texture of their regolith (Soderblom 1992; Roush et al. 1993; McKay 1996). Each body experienced differentiation soon after it was formed, followed by an extended period of regolith formation primarily from repeated impacts during the late heavy bombardment period of solar system evolution (3.8–4.1 Gyr ago; Tera et al. 1974). Photons hitting the regolith surface are both scattered from grain boundaries and absorbed and reradiated after interacting with the minerals in the soil (Burns 1993; Hapke 2005).

In the case of the Moon, the regolith contains minerals, glass and nano-phase iron (Pieters et al. 2000). The steep slope toward the NIR in the lunar data is primarily due to the continuously varying absorption coefficient of nano-phase iron that is produced by reduction of Fe and vapor deposition resulting from micrometeorite bombardment of the lunar surface. A large portion of the observed region of the lunar far side encompasses the South Pole Aitken Basin, which is enriched in mafic minerals relative to the near side (Nakamura et al. 2009). Charge transfer interactions with Fe²⁺ on the surface are responsible for the 950 nm absorption in the lunar spectrum (Lucey et al. 1998). It is important to note here that the 950 nm absorption cannot be due to the hydrated materials on the surface of the Moon detected by Sunshine et al. (2009) using the DI Medium Resolution Instrument (MRI) and the high-resolution near-IR spectrometer. The amount of lunar water is too small for its absorption effects to be measurable in the low spectral resolution HRI filter data.

In the case of Mars, the surface likely consists of fine-grained loess-like material as discussed in Singer et al. (1982). Surface mineral components containing optically active absorption bands in the visible are a combination of charge transfer absorption and spin-forbidden crystal field transitions due to Fe³⁺ and Fe²⁺ cations in ilmenite and pyroxene minerals. If the linear nature of the UV absorption in the lunar spectrum is indicative of the presence of nano-phase iron, it would appear that the disk-integrated Mars spectrum is not influenced by nano-phase iron. It is important to note here that the large vertical bars placed on the Mars data at longer wavelengths in Figure 5 do not reflect errors within the measurement, but instead reflect variations in the reflectance of the Martian surface observed during one full rotation. Irvine et al. (1968a, 1968b) also collected photometric data of Mars at various phases and rotational angles over a two-year time period (see Section 4.3). His spectrum of Mars at a 367 phase angle falls within the range of values observed with the HRI and is therefore consistent with our data.

Zoom In Zoom Out Reset image size

The disk-integrated reflectance spectrum of Earth shown in Figure 5 exhibits a steep incline toward UV wavelengths and is relatively flat at green to NIR wavelengths. The flat shape of the spectrum at longer wavelengths is attributed to clouds. Although, a slight rise in Earth's reflectivity from the 650 nm filter to the 850 nm filter is due to the fact that many surfaces on Earth (e.g., soils, sands, and vegetation) are more reflective at NIR wavelengths than at visible wavelengths. An absorption feature can be seen in the 950 nm filter due to a water overtone band located within this wavelength range (Traub 2003).

The increase in Earth's reflectance toward the UV is the result of Rayleigh scattering in the atmosphere. To investigate the significance of Rayleigh scattering in our Earth data, we produced two simulated high-resolution Earth spectra using the NASA Astrobiology Institute's Virtual Planetary Laboratory (VPL) three-dimensional spectral Earth model (Robinson et al. 2010) and convolved them with the HRI filter transmission curves, see Figure 6. In the first spectrum (solid line), realistic Rayleigh scattering is assumed and a rise in reflectivity at blue wavelengths is reproduced. The model used to produce the second spectrum (dashed line) is identical to the first except that Rayleigh scattering has been removed from the radiative transfer calculations. A strong rise in reflectivity toward blue wavelengths is no longer seen, although a slight increase is attributed to the blue color of Earth's oceans. These models agree with our assumption that the high near-UV reflectance of Earth is primarily due to Rayleigh scattering.

Zoom In Zoom Out Reset image size

Traub (2003) discusses the prospect of detecting vegetation on a planetary body by looking for the vegetation red edge (VRE), which he was unable to detect in his analysis of Earthshine data. The VRE is defined as

$$\begin{equation} {\rm VRE} = \frac{{r_{{\rm IR}} - r_{\rm R} }}{{r_{\rm R} }}, \end{equation} \tag{ 1 }$$

where r_IR and r_R are the reflectivities at NIR and red wavelengths, respectively. Narrowband filters designed to bound the steep rise in the reflectivity of planets around 700 nm are usually employed to measure the VRE. For example, a pair of 10 nm wide filters centered at 675 nm and 775 nm give a VRE value of ∼100% for a typical reflectance spectrum of a planet, ∼5% for a high-resolution reflectance spectrum of Mars, and between 0% and 10% for Earth (Arnold 2008), where the vegetation signature is washed out by other features on the planet (e.g., clouds). Cowan et al. (2009) used the EPOXI data set to correlate reddening in Earth's disk-integrated reflectance spectrum with the presence of landmass on the visible, illuminated disk. Vegetation-covered land was not distinguished from other land types because the medium-band HRI filters are nonideal for characterizing a steep rise in reflectivity and cannot properly distinguish between a gradual rise in reflectivity toward red wavelengths (as is the case for Mars, Mercury, and the Moon) and a steep rise in reflectivity near 700 nm, which is seen in some high-resolution, disk-integrated Earth spectra (Arnold et al. 2002).

To determine how the EPOCh data compare with the reflectance spectra of other solar system bodies, we expanded our analysis to include other terrestrial and Jovian worlds. Observations for these bodies were not collected with the DI HRI, so full-disk reflectance data from previously published studies were used for our analysis. We reproduced reflectance spectra of Mercury and Venus from Irvine et al. (1968a, 1968b) collected at Le Houga Observatory, France and Boyden Observatory, South Africa. The observations were taken through 10 medium-band filters centered between 315 nm and 1060 nm, and the calculated magnitudes of the planets were normalized to a distance of 1 AU from the Sun. We chose the observations with phase angles closest to that of the May EPOCh observations of 75°; the angles for the Mercury and Venus were 77° and 76°, respectively. The band centers of the filters used by Irvine et al. are similar to the HRI filters, so a linear interpolation and conversion to reflectance was necessary for comparison with the EPOCh data. Figure 7 shows a comparison between the EPOCh and Irvine data sets.

Zoom In Zoom Out Reset image size

The terrestrial worlds, other than Earth, are all less reflective in the UV than in the NIR. Mercury, an airless body, has a reflectance spectrum increasing steadily from the UV to the IR due to light interactions with surface minerals (Vilas 1985). The Mercurian soil is low in Fe with <2–3 wt % at its surface, so Fe is not a major contributor to charge transfer slope in the reflectance of Mercury (McClintock et al. 2008). The visible colors of Mercury are not diagnostic of the surface mineral composition.

The reflectivity of Venus shows a depression in the UV and a relatively flat reflectance from the green to the NIR filters. Venus has a thick atmosphere unlike the airless bodies discussed previously, so its atmospheric structure and composition are solely responsible for the planet's colors. Venus' opaque atmosphere contains several distinct layers of sulfuric acid clouds, which are distinguished by their aerosol particle size. The cloud deck which is primarily responsible for Venus' high reflectivity is located between about 50 and 80 km in altitude (Marov et al. 1973; Marov 1978). Extinction optical depths through the Venusian haze (which extends up to about 90 km in altitude) and the upper cloud deck reach unity by about 30 mbar (Knollenberg et al. 1980), so no appreciable amount of Rayleigh scattering can occur. The effects of a "UV absorber" can be seen in the reflectance spectrum of Venus blueward of about 650 nm. The absorption efficiency of this substance increases from 650 to 300 nm, explaining the drop in Venus' reflectivity at blue wavelengths (Pollack et al. 1980). The flat reflectance of Venus' reflectivity curve beyond about 650 nm is explained by the sulfuric acid droplets' essentially non-absorptive behavior at visible wavelengths (Crisp 1986).

In addition to Irvine's photometric data, we used high spectral resolution, full-disk reflectance data for Jupiter, Saturn, Uranus, Neptune, and Titan from Karkoschka (1994). The observations were collected at the European Southern Observatory and span the wavelength range of 300–1000 nm at phase angles of 98, 27, 06, 04, and 27, respectively. We convolved these spectra with the HRI filter transmission curves to determine the medium-band reflectance spectra of the five bodies. An overlay of the high-resolution reflectance spectrum of Jupiter from Karkoschka (1994) and the HRI filter transmission curves are seen in Figure 8, and the computed broadband photometric reflectance values for the Jovian worlds are shown in Figure 9.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Uranus and Neptune are unique in that their reflectance spectra are relatively flat at blue and UV wavelengths while containing strong absorption features in the red and NIR. A perfectly scattering, semi-infinite atmosphere consisting of pure H₂ would have a wavelength-independent, zero-phase reflectance of about 0.8 (Prather 1974), which is about 30%–40% larger than the reflectance values reported by Karkoschka (1994, 1998) for Uranus and Neptune at short wavelengths. The discrepancy between the theoretical value and the measured value is due to Raman scattering, which decreases the zero-phase reflectance values for these worlds by about 5%, and the presence of aerosols (Karkoschka 1994). In the case of Uranus, a methane aerosol haze near 1.3 bar and a thick cloud of H₂S aerosols significantly reduce the reflectance values at UV and near-UV wavelengths (Baines & Bergstralh 1986). The methane haze preferentially absorbs blue wavelengths, which is possibly due to contamination from a thin stratospheric haze (Rages et al. 1991), leading to a drop in Uranus' reflectance between the 550 nm and 350 nm filters. Similarly for Neptune, a methane haze near 1 bar in the atmosphere and photochemically produced stratospheric hazes of hydrocarbons reduce the reflectance values at blue and UV wavelengths (Baines & Smith 1990). Both Uranus and Neptune experience strong CH₄ absorption features at wavelengths longer than about 500 nm (Karkoschka 1998). The wings of these absorption features deplete most of the continuum at red and NIR wavelengths, and are especially strong in the 850 nm and 950 nm HRI filters. The single-scattering albedos of aerosols in the atmospheres of Uranus and Neptune decrease between green and red wavelengths, causing a slight drop in the continuum reflectivity of these worlds at red wavelengths (Baines & Bergstralh 1986; Baines & Smith 1990).

The reflectance spectra of Jupiter, Saturn, and Titan are different from the ice giants as they exhibit weaker absorption in the NIR and stronger aerosol absorption at blue and UV wavelengths. High-resolution reflectance spectra of Jupiter and Saturn are similar in appearance. Both are dominated by methane absorption at wavelengths longer than about 700 nm, especially in the 850 nm and 950 nm HRI filters (Karkoschka 1994, 1998). At shorter wavelengths, the reflectance spectra of these worlds begin to decrease due to aerosol absorption. The composition of these aerosols is currently unknown, but is commonly referred to as "chromophores." Compounds that may explain the blue absorption in the atmospheres of Jupiter and Saturn include photochemically produced hydrocarbons and/or sulfur-bearing compounds (see West 2007; Simon-Miller et al. 2001). Models predict multiple cloud decks in the atmospheres of Jupiter and Saturn (Weidenschilling & Lewis 1973), which was verified in the case of Jupiter by the Galileo probe (Ragent et al. 1998). The uppermost cloud deck of these worlds is composed of ammonia ice, which tends to form at higher pressures (lower altitudes) in the Saturnian atmosphere due to Saturn's colder temperature profile (Atreya et al. 1999; Gierasch & Conrath 1993). A slight upturn in reflectance at UV wavelengths for Saturn is attributed to Rayleigh scattering (McCord et al. 1971). The spectral behavior of Saturn in this wavelength region is controlled by competition between aerosol absorption and Rayleigh scattering.

Titan's reflectance spectrum is similar to that of Jupiter and Saturn, and exhibits absorption at both UV and NIR regions. At visible wavelengths, Titan's reflectance spectrum is dominated by absorption from a stratospheric haze at short wavelengths, scattering and absorption by nitrogen-rich hydrocarbon aerosols (tholins) at intermediate wavelengths, and gas absorption at longer wavelengths. Spectroscopic and laboratory studies have revealed that the tholins which make up Titan's haze generally have an increasing absorption efficiency with decreasing wavelength through the visible (Rages & Pollack 1980; McKay 1996; Tomasko et al. 2008). As a result, Titan's reflectance spectrum decreases steadily from 650 nm to UV wavelengths. Methane absorption dominates Titan's reflectance spectrum longward of 650 nm. Although these features are not as strong as those in the spectra of Uranus and Neptune, the 850 nm and 950 nm HRI filters show marked decreases in Titan's reflectance due to methane absorption.