4.1. Patch Spectral Properties
 The high-albedo patches all have pronounced absorptions at 0.967 μm. The 0.967 μm filter is prone to artifacts, because it was the final filter taken during the sequence of images, and illumination conditions can change. However, we have used several criteria to rule out the possibility of artifacts. First, the pixels are clumped together, not randomly distributed across an image. Second, they are not located in or near shadows. Third, they were observed repeatedly and have the same feature on multiple sols (up to 113 sols). Fourth, these pixels have the same feature in images taken at multiple phase angles (Figure 3), ruling out the possibility of a viewing geometry artifact. Fifth, each image had a slightly different pointing, so the pixels in question are not confined to specific parts of the detector.
 We conclude that the 0.967 μm feature is not an artifact, but rather is caused by a material exposed during RA operations. Additionally, because the feature is found only as patches on the clods, we conclude that the material is a surface coating or crust.
4.2. Patch Spectral Candidates
 Water ice, which is also observed in trenches, has absorptions at 0.98 and 1.04 μm, which cause a negative slope from 0.85 to 1.001 μm. Water ice is a poor spectral fit to spectra of the patches described here, thus is an unlikely candidate. Additionally, the long-term stability of these patches rules out water ice, which was observed to sublimate in a matter of hours to sols when exposed at the surface [Smith et al., 2009].
 Several minerals (e.g., Figure 3c) produce a 0.967 μm feature, including hydrated Mg- and Ca-perchlorates, some zeolite and phyllosilicate minerals, and at least one hydrated chloride mineral (bischofite [Crowley, 1991]). WCL could detect and measure Cl ions, and did not detect bischofite [Kounaves et al., 2010a, 2010b]. Our geochemical modeling indicates that zeolites are unlikely candidates for the patches described here, because they are unlikely to form under Phoenix site conditions (see Section 3.1 above). Zeolite formation generally requires an aqueous environment with pH > 9 [Sheppard and Hay, 2001], and though the Phoenix landing site may have experienced aqueous activity in its past [Smith et al., 2009], the 7.7 pH of the soil appears to be buffered by calcite [Boynton et al., 2009], making a highly alkaline aqueous environment unlikely.
 In one soil sample, Phoenix's Thermal and Evolved Gas Analyzer (TEGA) recorded a very small H2O release between 700 and 800°C that was interpreted as possible evidence of a phyllosilicate mineral (possibly smectites [Boynton et al., 2008]). However, material discussed here is present only in concentrated subsurface patches, indicating that some mechanism is translocating them down in the soil column, and concentrating them into patches. Phyllosilicate minerals have a very low solubility and would not be expected to be redistributed by dissolution and reprecipitation in water. Aqueous clay illuviation can physically move phyllosilicates from the surface to subsurface, forming argillic horizons or clay coatings on peds; however, this process requires repeated flushings with a substantial quantity of water [e.g., Eswaran and Sys, 1979]. Additionally, Phoenix soils undergo pedoturbation resulting from seasonal freeze-thaw cycles [Mellon et al., 2008]; hence, the high-volume wetting events would have had to occurred recently in order for the argillic horizons to remain intact. We see no evidence to suggest that the Phoenix landing site has been repeatedly flushed with large volumes of liquid water in the geologically recent past, and so conclude that physically translocated phyllosilicates are unlikely candidates for the patches reported here.
 Additionally, our geochemical modeling suggests that the stable phyllosilicates would all contain substantial iron because of their formation from basaltic material, and they would display a green or brown color, which is inconsistent with the observed crust. Their original formation would also require substantial aqueous alteration of basaltic sand; corroborating evidence for such alteration at the site is lacking.
 Perchlorate, on the other hand, is highly soluble in even very small amounts of water, and would be easily transported from the surface to the subsurface by fluids. In the subsurface, it would form concentrated crusts as the water evaporated or froze then sublimated. Given the previous WCL detection of perchlorate at the site, the thermodynamic instability of zeolites, the concentrated morphology of the observed patches, and their location in the soil column, we conclude that a hydrated perchlorate salt is the most likely candidate to explain the 0.967 μm absorption feature.
4.3. Perchlorate at the Phoenix Landing Site
 The observation that concentrated perchlorate patches are limited to the subsurface is not inconsistent with previous findings from the Phoenix WCL, which reported perchlorate throughout the soil column, including a sample from near the surface (actually a scoop sample of the upper ∼1 cm of soil) [Hecht et al., 2009]. Taken together, the WCL and SSI observations indicate that the soil column contains low concentrations of evenly dispersed perchlorate, with occasional patches of the highly concentrated perchlorate reported here.
 These observations have constrained the cation associated with the perchlorate: only Mg- and Ca-perchlorate produce a 0.967 μm feature, while K-, Na-, Fe3+, and Fe2+− perchlorates do not. Interestingly, Mg- and Ca-perchlorates have significantly lower eutectic temperatures, increasing the chances to brine formation [e.g., Chevrier et al., 2009]. This work does not rule out the possibility of additional perchlorate phases at the Phoenix site.
 Previous studies have proposed that perchlorate on Mars might form from atmospheric interactions between ozone and volatile chlorine compounds as aerosols or at the surface, as in the Atacama Desert of Chile [Catling et al., 2010] and the Antarctic Dry Valleys [Kounaves et al., 2010a, 2010b]. Dissolution and redistribution of perchlorate could occur when summer mid-day temperatures exceed the perchlorate eutectic point. Chevrier et al.  showed that Mg-perchlorate is metastable above 206K, while Phoenix surface temperatures rose up to 245K during the mission. However, the volume of water in the atmosphere during these times is small: the maximum observed water vapor in the atmospheric column was ∼55 precipitable- μm [Tamppari et al., 2009]. An alternative redistribution mechanism involves seasonal ices. The Phoenix site is covered in early fall by an ∼90 micrometer layer of seasonal H2O ice, which is then topped in winter by a translucent slab of CO2 ice that reaches ∼30 cm thick [Cull et al., 2010]. At the base of the translucent CO2 slab, a solid greenhouse effect can increase the temperature at the ice-surface interface, a process that Kieffer et al.  proposed could be responsible for southern hemisphere “spider” features. We propose that this solid greenhouse effect can raise temperatures at the ice-surface interface high enough for small amounts of meltwater or thin water films to form, dissolving surface perchlorate and transporting it downward through diffusive or gravity-driven fluid transport. Transport into the soil is limited by the rate of transport and the lower boundary of the subsurface ice table. This surface-to-subsurface redistribution process is common in the Antarctic Dry Valleys, where it concentrates soluble sulfates beneath soil clods and rocks. Perchlorate is thus removed from the surface and deposited as salt crusts in the shallow subsurface soil by thin films of water (greater volumes of liquid water are unlikely given the small amount of seasonal water ice involved and the low relative humidity/partial pressure of water in the atmosphere). This scenario implies a geologically recent occurrence of aqueous processes at the site.