Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 We report on mapping of the north polar region of Mars using data from the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) instrument. We have observed 3 Mars years (28–30) of late winter and spring recessions (Ls = 304°–92°). Our investigations have led to the following observations. (1) We classify the retreat of the north polar seasonal cap into “presublimation,” “early spring,” “asymmetric” and “stable” periods according to the prevalent H2O ice grain size distributions. (2) During the early spring, the signatures of CO2 ice at the edge of the cap are obscured by H2O ice, which increases the apparent size of the H2O ice annulus around the seasonal CO2 cap at this time. At around Ls = 25°, this process changes into an asymmetrical distribution of H2O deposition, covering CO2 signatures more rapidly in the longitude range from 90 to 210°E. (3) We detect signatures of “pure” CO2 ice in extremely limited locations (in Lomonosov Crater) even in midwinter. H2O ice signatures appear everywhere in the retreating CO2 seasonal cap, in contrast with the south polar seasonal cap. (4) We find that average H2O ice grain sizes continuously increase from northern midwinter to the end of springtime; this is the inverse of the behavior of CO2 ice grain sizes in the southern springtime.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Martian polar regions are key to understanding the current climate and energy balance of the Red Planet. More than 25% of the Martian atmosphere participates in the polar sublimation and condensation cycle [Hess et al., 1977, 1979, 1980; Kelly et al., 2006]. This dynamic movement of carbon dioxide between the atmosphere and the surface is important to understand since it has implications for past, present and future Martian habitability. Even as we study this process in greater detail, we now have information at hand that suggests the cycle itself may be changing [Haberle and Kahre, 2010] and that reservoirs of CO2 ice are sequestered in the polar regions and not currently interacting with the Martian seasonal cycle [Phillips et al., 2011].
 This study reports the seasonal changes in the north polar region of Mars as observed by the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) and the Mars Color Imager (MARCI) camera data on the Mars Reconnaissance Orbiter (MRO) spacecraft during its first 3 Mars years in orbit. Mars years are abbreviated MY and MY 1 started 11 April 1955 [Clancy et al., 2000]; here we report on observations from MY 28–30. An accompanying study of the south polar region was recently reported in Brown et al.  and a similar study has been completed on a smaller scale at Louth Crater [Brown et al., 2008a].
 The CRISM instrument is a visible to near infrared imaging spectrometer with spectral coverage of the ∼0.36–3.92 μm range [Murchie et al., 2007]. The S-channel detector on CRISM covers the 0.362–1.053μm range and the L-channel from 1.002 to 3.92μm. We used only L-channel spectra in this study.
 CRISM operates with a gimbal to obtain full and half resolution images suitable for geological mapping at ∼18 m/pixel. In addition, CRISM can operate in a nadir pointing “mapping mode” to collect 10x binned pixels (∼182.5 m/pixel) with a smaller number of bands for reduced data rate suitable for global mapping. The observations reported here are all taken in CRISM mapping mode. The CRISM swath width on the ground is a narrow ∼10 km, which is responsible for the “spaghetti strand” appearance of maps in the polar region (Figure 1).
 The MARCI instrument is a wide-angle 180 field of view camera that images Mars every day with 12 continuous limb to limb color images, each covering 60 of longitude. MARCI has 2 UV channels and 5 visible channels. The visible channels used in this paper have 1 km resolution [Malin et al., 2001].
2.2. CRISM Seasonal Mosaics
 For this study, we have collected all mapping observations acquired north of 55° latitude and constructed seasonal mosaics of observations based on the two week MRO planning sequence using the MR PRISM software suite [Brown and Storrie-Lombardi, 2006]. All maps are presented in a polar stereographic projection from 55°N to the pole (Figure 1). Mosaics were constructed pixel by pixel, and overlapping data was overwritten by the last image acquired during the period. No averaging of pixels was attempted.
 CRISM pixels were processed with a “cos(i)” correction using the incidence angles supplied with the CRISM image backplane data based on MOLA topographical data. No atmospheric correction has been attempted. All albedos discussed in this paper are therefore apparent Lambert albedos, which assume isotropic scattering of incident light back to the observer. More information on CRISM processing is presented in Brown et al. .
 CRISM collects mapping data in two forms: CRISM MSP (Multispectral Polar, which is a multispectral mapping mode) and HSP (Hyperspectral Polar) images. These are collected in nadir mode and have a reduced set of bands to minimize data rates. MSP observations have 19 “S” (short wavelength) bands and 55 “L” channel bands, HSP observations have 107 S channel bands and 154 L channel bands.
 CRISM commenced imaging of the north polar region in September 2006, in the 2 week transition phase prior to commencement of the MRO science mission phase. MRO suffered unexplained resets of the onboard command computer during 2009 and was placed in maintenance mode from 26 August to 16 December 2009 (late northern winter MY 29 to early northern spring MY 30), during which time no CRISM data was acquired. During September 2006 to January 2010, only MSP-type mapping observations were collected. In MY 30 (13 January 2010) HSP-type mapping observations were added in response to unanticipated extra MRO-downlink capability. After 13 January 2010, the CRISM instrument collected HSP and/or MSP mapping data as determined by the JHU/APL science team based on variable data downlink capacity.
 A summary of north polar mapping observations for each MRO planning period is presented in Table 1.
Table 1. Number of CRISM MSP Strips Taken of Mars Northern Pole (Defined as All Strips Partially or Completely Poleward of 55°N) per Fortnight (Starting at the Transition Cycle of the MRO Primary Science Mission)a
Northern spring starts at Ls = 0° and ends at Ls = 90°. MSP, multispectral polar observation; HSP, hyperspectral polar; MY, Mars year; Ls, solar longitude from Mars; DOY, Earth day of year. Bold text indicates summer observations, which are not specifically discussed in this paper.
MY values are given in parentheses.
No North Observations During Northern Fall (24 Mar to 5 Aug 2007)
No North Observations During Northern Fall/Winter (15 Feb to 11 Jul 2009)
CRISM off Due to MRO Maintenance to 26 Aug to 16 Dec 2009
2.3. H2O Ice Detection and Grain Size Estimation Strategy
2.3.1. H2O Ice Detection
 H2O ice dominates the north polar cap, and to map its presence in a CRISM pixel we use a H2O ice index first proposed by Langevin et al.  and modified for CRISM mapping data by Brown et al. :
where R(λ) is the apparent reflectance at wavelength λ in microns.
 In our ice identification maps, we decide that H2O ice is present if the H2O index is above a threshold value of 0.125, which we determined by iterative assessments of noise removal from our maps.
2.3.2. H2O Ice Grain Size Estimation
 In order to estimate the water ice grain size from the H2O ice index, we used a monodisperse, one dimensional approximate radiative transfer model to construct artificial H2O ice reflectance spectrum proposed by Shkuratov et al. . This model has a “porosity” parameter lying between (0,1] that describes the volume fraction occupied by the target material and free space. A porosity value of 1 indicates no free space and a value of 0.5 indicates that half the volume is occupied by the target material.
 The Shkuratov model produces as output a 1D grain size parameter that may be imagined as an “equivalent path length” for photons between scatterings, therefore the grain sizes we quote in this paper are actually fictionalized average photon scattering path lengths. These do not correspond to a spherical or cylindrical diameter or radius because the model is 1 dimensional and does not replicate spherical shapes. Therefore in this paper, the term “grain size” is the same as the equivalent path length of light within the model.
 We used water ice optical constants of Warren  and palagonite optical constants of Roush et al.  (as a dust simulant) as input to the Shkuratov model. After carrying out the calculations at the resolution of the optical constants, we then convolved the model output spectrum to CRISM wavelengths before measuring the H2O index of the model spectra.
 The dust grain size and volume percentage were varied in order to keep the visible albedo of the mixture around 0.5, which is typical of residual ice cap apparent visible albedos [Kieffer, 1990]. Here we consider “visible albedo” to be equivalent to the apparent Lambert reflectance at ∼0.8 μm. In order to match this albedo, we iteratively adjusted the volumes of our two-component model of H2O ice and dust. We found that a reasonable model resulted from H2O ice occupying 70% of the volume and dust grains (with a size of 35 microns) occupying 30% of the volume.
2.3.3. Model to Convert H2O Index to Water Ice Grain Size
Figure 2a presents the curve for conversion of H2O index to water ice grain size we have used in this study. We have chosen to report the H2O index in our maps and histograms rather than grain size because grain size derivations require assumptions about the snowpack porosity, size distribution and an estimate of other components (e.g., dust). We consider it important to present estimated grain sizes, however there are caveats and assumptions in any model that infers grain sizes from reflectance spectra and we detail these in the paragraphs below.
 To generate Figure 2a, we created the simplest spectral model we could imagine that can roughly fit the CRISM spectra observed. It is possible to imagine more complex scenarios, however our attitude to fitting infrared spectra follows the two principles of (1) Occam's Razor and (2) “if you do not observe a unique set of absorption bands, make sure you clearly explain your deductive reasoning.” To this end, we used our two component H2O ice/dust model in an attempt to simulate a H2O ice snowpack with embedded (intimately mixed) dust grains. This two-component model is clearly an oversimplification of reality but we believe that it is necessary to adopt such a model in order to account for the relatively low visible albedo of the cap (pure H2O ice typically exhibits visible albedos of 0.99 [Grenfell et al., 1994] whereas, as stated earlier, the north polar cap typically exhibits visible albedos of 0.5).
 Our approach has the advantage of calculating the reasonably straightforward “minimum apparent grain size” [Brown et al., 2009]. Because we have chosen a two component (ice-dust) model, rather than a range of components or a size distribution, we are essentially calculating the smallest water ice grain size that will fit the observations. This means that we have made a positive detection of the minimum water ice grain size that is required to explain the observations; we can say with certainty that water ice of at least this grain size is present, otherwise the 1.5μm water ice feature would not display this band depth. Use of the minimum apparent grain size phrasing is not common in the radiative transfer literature, but we feel this terminology merits wider adoption given its simplicity and physical plausibility [Brown et al., 2009].
2.3.4. Model for Water Ice Grain Sizes Larger Than 100 Microns
 Beyond 100 microns grain size, the depth of the 1.5 μm band for our two-component model becomes saturated and it is therefore not suitable for grain size retrieval. For this reason, beyond a H2O index of ∼0.5, if one is to obtain reliable grain size estimates, it is necessary to use an alternate water ice index. We have chosen to use the depth of the 1.25 μm band, which becomes prominent (greater than 0.02 relative band depth) for grain sizes larger than 100 microns. Figure 2b shows the curve for converting between 1.25 μm band depth and grain size.
 To generate Figure 2b, we used the Shkuratov model described earlier (with a mixture of 70/30 percent by volume water ice/palagonite) to deduce the relationship between the 1.25 μm band depth and the water ice grain size. To estimate the amplitude of the 1.25 μm band depth, we used a band depth algorithm described in Brown  and implemented in Brown et al. [2008b]. We discuss the implications of using the 1.25 μm band depth in this manner in sections 3 and 4.
2.3.5. Errors Induced by H2O Snowpack Density and Porosity Estimation
 In order to make an estimate of the errors induced by assuming a porosity of 0.5 in our spectral model, we have constructed models where porosity is 0.1 (fluffy snow), 0.5 and 1.0 (slab ice). These three models are shown in Figure 2c. Using our assumed porosity of 0.5, for H2O ice with a grain size of 10 microns, the range of induced error is ±1 microns, for grain sizes of 100 microns, the range is ±6 microns and for grain size of 250 microns, the range is ±12 microns. Based on these ranges, we can make an assessment that there is a relatively large induced error (up to ±6%) introduced if our porosity estimate is incorrect. The uncertainty introduced by noise is minimal, since CRISM SNR at 1.5 μm is ∼400 [Murchie et al., 2007]. For a band depth of 0.08 this would equate to (1/400)/0.08 = 3.5%, almost half of our porosity error. Assuming the porosity and instrument errors are statistically independent, and using the relation total error = sqrt(errorA2 + errorB2) then the instrument error gives an additional ∼0.6% to the porosity error. Rounding up to one significant digit, we use ±10% for the relevant error estimate when quoting grain sizes later in the text.
2.4. CO2 Ice Detection Strategy
 In order to detect and map CO2 ice we used the CO2 absorption band centered at 1.435 μm. To construct our ice identification maps, we determined that CO2 ice was present if the 1.435 μm band depth exceeds a threshold value of 0.16 [Brown et al., 2009]. We have not attempted to derive maps or histograms grain size of seasonal CO2 ice in this paper due to the challenges of separating H2O and CO2 ice signatures (there is almost no “pure” CO2 ice in the maps we created here, in contrast with the south [Brown et al., 2009]). We report on changes of spectra of CO2 ice at individual locations in section 3.
2.5. Seasonal Cap Retreat Line Mapping
 As discussed in detail in Brown et al. , we have devised an automated method to find the edge of the polar cap. In this paper, the “Cap Recession Observations indicated CO2/H2O has Ultimately Sublimated” (CROCUS/CROHUS) line corresponds to the most equatorward detection of CO2/H2O ice. The term “CROCUS” was first introduced by Kieffer et al.  to describe the retreating edge of the CO2 seasonal cap, the term CROHUS is introduced here to describe the edge of the H2O ice seasonal cap. For comparison with previous observations utilizing only visible imagery to detect the edge of the cap, the edge of the visible cap corresponds to the CROHUS (H2O) line and the CROCUS (CO2) line lies poleward of this (since H2O ice is more stable than CO2 ice to ambient temperature changes). This distinction is discussed in greater detail below.
 We have constructed MARCI daily global mosaics in a similar manner to Brown et al. . For an in depth analysis of MARCI cloud and dust observations over a similar time period as this study, see Cantor et al. .
3.1. Seasonal Mosaics
 The seasonal CRISM and MARCI mosaics are presented in Figures 3a–3e. Alongside each image is the Ls range of the CRISM observations, the number of CRISM mapping images used to construct the mosaic, and the Ls corresponding to the MARCI image.
 For Figures 3a–3d, the left column displays a MARCI image, the central column displays the CO2ice-related 1.435μm band depth and the right column displays the ice identification map for each time period.
Figure 3e differs from Figures 3a–3d because instead of the 1.435 μm band depth maps in the central column, the H2O ice index (equation (1)) has been mapped. This is because no CO2 ice signatures are apparent during these time periods. Figure 3e also includes an extra column (on the far right), showing the 1.25 μm band depth maps for each period for comparison with the H2O ice index.
Figure 4 highlights the dynamic evolution of CO2 and H2O ice by showing spectra from several different locations at different times in the cap recession. Figure 4a shows examples of coarse and fine grained CO2 ice. The coarse grained example is taken from a midwinter MSP observation of Utopia Planitia, and the fine grained example comes from the outer slope of Lomonosov Crater at the same period (they are ∼0.1° Ls apart). With a 1 μm albedo of around 0.85, the fine grained CO2 ice spectrum is the brightest CRISM observation in the north pole that we have observed. Using grain size estimation methods in Brown et al. [2009, Figure 2b] the fine grain CO2 ice has a grain size of 2 mm (2.28 μm band depth of 0.08) and the coarse grained CO2 ice spectrum from Utopia Planitia has a grain size of 10 mm (2.28 μm band depth of 0.22). We cannot rule out coarser or finer grained CO2being present in regions where CRISM did not collect observations (for example, the polar regions north of ∼70°N were not observed during the winter period due to low-light conditions).
Figures 4b and 4c present time evolution of polar cap spectra at two locations. Figure 4b presents observations from 45°E, 85°N on Gemini Lingula, “Point B” of Langevin et al. , who observed the water ice grain size growth from Ls = 93°–127°. This region is also featured in Figure 7 of Byrne et al. , where MOLA 1.064 μm albedos were obtained. This location is ideal for near-continuous observations due to the lack of topography and it is covered by ice year-round.Figure 4bshows two springtime MSP L-channel spectra demonstrating that Point B is covered by CO2 ice at Ls = 8° and Ls= 27° (continuum-removed band depths of the 1.435μm band are 0.45 and 0.39) then the CO2 ice disappears at Ls = 34° (1.435 μm band depth of 0.06). From Ls = 34°–76° the water ice absorption bands are remarkably stable (with a H2O ice index of ∼0.55 or around 50 microns from Figure 2a). Langevin et al.  reported water ice grain growth beginning around Ls = 93°; the spectra in Figure 4b show that ice grain growth actually commences around Ls = 89°. The significance of what we have shown here is that the process is one of stagnation in ice growth for most of spring, followed by a large increase in ice growth starting around Ls = 89°. This set of spectra therefore presents the strongest evidence for in situ thermal metamorphism of water ice in the Martian north pole to date.
Figure 4c presents the time evolution of CRISM spectra collected from 164.2°E, 72.4°N in Korolev crater, which is “Point C” of Langevin et al. . Korolev has been shown to contain isolated regions of high thermal inertia which correspond to topography and have been interpreted as mounds of ice-rich regolith [Armstrong et al., 2005, 2007].
 The spectra in Figure 4c show that CO2 ice is present in Korolev during Ls= 3° and 13° (continuum-removed band depths of the 1.435μm band are 0.31 and 0.33), and then around Ls = 46° the CO2 ice is almost completely gone (1.435 μm band depth of 0.20). The Ls = 46°, 65° and 89° observations show increases in the strength of the water ice bands. The H2O ice index for these spectra are 0.44, 0.55 and 0.56, respectively; the 1.25 μm band depths are 0.05, 0.04 and 0.14. This demonstrates that the 1.5 μm band has saturated between Ls = 65° and 89°, as can be seen by the flattening of the 1.5 μm band in Figure 4c. This absorption band behavior is likely caused by rapid water ice grain growth (contrast this with the almost identical depth of the H2O ice absorption bands from Ls = 34°–76° in Figure 4b). The faster increase in water ice absorption band growth of Korolev relative to Point B is congruent with our interpretation of thermal metamorphism; water ice in Korolev is exposed to a warmer climate that will drive faster grain growth [Gow, 1969].
 Comparison of the spectra at Point B and Point C demonstrates the inability of the H2O index (based on the 1.5 μm band depth) to discriminate effectively between water ice of grain sizes larger than ∼100 microns. At Ls = 89°, the Point B and C 1.5 μm band depths are very similar (0.58 and 0.56, respectively) however their 1.25 μm band depths are 0.04 and 0.14, demonstrating a true grain size of ∼100 (Point B) and ∼400 (Point C) microns according to Figure 2b.
3.2. Seasonal Ice Cap Boundaries and Cap Area
Figures 5a and 5b plot the CROHUS line and Figure 5c plots the CROCUS line (defined using our ice identification thresholds mentioned earlier) as a function of time. Due to our interpolation method that is required to skip areas of missing data, the edge of the cap is unrealistically simplified, however our results are qualitatively similar to observations on previous years. Figure 5d shows the CRISM observations of the edge of the CO2 ice cap and H2O ice cap from Ls = 0°–90° for MY 29 compared with observations for previous years [Kieffer and Titus, 2001; Titus, 2005; Wagstaff et al., 2008; Appéré et al., 2011]. CRISM CROCUS/CROHUS boundaries are given as ranges covered by a box in order to illustrate the range of minimum and maximum latitude of the cap edge. It can be seen that the CO2 cap edge boxes are very wide during the asymmetric retraction period. The CRISM cap edges have been constructed automatically and we believe this accounts for the slightly faster retreat rates when compared to other observations, otherwise the comparisons show the cap retreat is relatively similar to other years.
 In order to calculate the area of the cap, we used the polygons in Figure 5, and applied an algorithm for computing the area of an irregular polygon available in the open source Computational Geometry Algorithms Library (CGAL). We show the estimate of the cap area in Figure 6. It can be seen that the cap covers just less than 8% of the Martian surface at the midwinter period and decreases to its summertime area around Ls = 90°, when it covers ∼1.4% of the Martian surface.
3.3. Histograms of Seasonal H2O Ice Grain Size
 The histograms of seasonal water ice abundances for each set of CRISM observations are presented in Figures 7a–7e. Figures 7a–7d use the H2O index and Figure 7e uses the 1.25 μm absorption band depth. The histograms have 100 bins and are normalized by the number of observations containing H2O ice. In Figures 7a–7d, all pixels were included in the histograms. In Figure 7e, all pixels having a H2O index of greater than 0.125 were included in the histograms, even if those pixels also contained CO2 ice. The histograms show that H2O ice indexes increase throughout the springtime, allowing us to divide the cap retreat into several time periods as discussed below.
4.1. Interpretation of Seasonal Mosaics
 Examination of the CRISM seasonal maps leads to a natural division of the springtime recession into four different phases, presublimation, early spring, asymmetric retraction, and stable phases. The first phase, from Ls = 304°–0° we term presublimation phase. From Ls = 0°–25°, is the early spring phase, from Ls = 25°–62° is the cap asymmetric retraction phase and from Ls = 62°–92° is the stable cap phase.
4.1.1. Apparent CO2 Ice Cap
 In the following discussion we refer to the “apparent CO2 ice cap” in order to indicate those areas where CRISM detects CO2 ice (according to the definitions above). As discussed further below, that does not match the TES CO2 ice temperature map [Kieffer and Titus, 2001], which extends equatorward of the CRISM apparent CO2 ice cap, indicating that H2O ice (most likely transported and cold trapped by the Houben process) is obscuring a layer of underlying seasonal CO2 ice that is responsible for the observed TES temperatures [Wagstaff et al., 2008; Appéré et al., 2011].
4.1.2. Presublimation Phase (Ls = 304°–0°)
Figures 3a and 3b cover this phase. This phase embodies the winter state of the Martian northern pole. The seasonal cap clearly extends equatorward of 50°N and is roughly symmetric around the pole at Ls = 304°. Around Ls = 330°, the retreat of the apparent CO2 cap seems to slow along the 0°E line. By Ls = 340°, the retreat is roughly symmetric again, and the apparent CO2 ice cap edge is close to 58°N. The MARCI images show the seasonal cap has relatively low visible albedo during this time.
220.127.116.11. Edge of the Apparent CO2 Ice Cap Signatures
 At the edge of the CO2 (mixed with H2O) seasonal cap, we observe a decrease in the strength of the 1.435 μm CO2 absorption band, as shown in the middle column of Figures 3a and 3b by blue and green pixels at the edge of the cap in the middle column. This decrease in 1.435 μm CO2 band strength at the edge of the cap could be due to (1) a decrease in area covered by the CO2 cap (at fractions of the ∼187.5 m scale), (2) obscuration of the CO2 ice by H2O ice, or (3) a decrease in the grain size as the edge of the cap is sublimating. These three effects are not separable without high-resolution images of the cap edge, and in fact they may all act in concert to produce the effect observed. The edge of the cap thus affected by this decrease in the CO2 ice absorption band forms an annulus around 10–100 km wide around the retreating cap.
18.104.22.168. Pure CO2 Surface Ice Deposits in Lomonosov Crater Between Ls = 304°–336°
 In contrast to the south pole, where pure CO2 ice deposits (i.e., no H2O ice absorption bands present in a spectrum) are common, Figure 3a shows that it is extremely rare to find pure CO2 ice in the north pole, even in winter time observations. The exception is in Lomonosov crater at (at 65°N, 350°E) where pure CO2 ice is present from Ls = 304°–336° (see also Figure 4a). This is the only region that CRISM detected pure CO2ice over this middle-to-late winter time, although observations poleward of 70°N were not carried out in midwinter due to low-light conditions. The fact that the pure CO2 observations were sharply limited to the crater and present for such an extended period suggests that the crater may enjoy its own microclimate during this period which inhibits H2O ice deposition (or encourages CO2 ice deposition).
22.214.171.124. Decreasing Width of H2O Ice Annulus
 At Ls = 304°–311°, the earliest observations we have available, the H2O ice annulus can be seen to extend to (Figure 3a, third column) 47°E, 47°N. We checked a CRISM MSP observation taken at Ls = 309°, (MSP 787D_01), that is not in Figure 3a, but extends further south. This observation shows that the true extent of the water ice annulus extends to 48°E, 44.3°N at Ls = 309°. The CO2 ice signature extends at this time to 46°E, 55°N, thus the water ice annulus spans over 10.7° of latitude at this time.
4.1.3. Early Spring Phase (Ls = 0°–25°)
Figures 3b and 3c cover this phase. The middle columns of these images clearly show the rapid retreat of the CO2 ice signatures from ∼62°N to ∼70°N. The exterior edge of the seasonal cap, as estimated by CRISM H2O ice signatures, retreat far more slowly, from ∼59°N to 62°N, during the same period.
 The seasonal cap CO2ice distribution during the Early Spring phase has an inhomogenous distribution, particularly apparent during the well-covered two week period of MY 29 Ls = 4°–16°. Deeper CO2ice-related 1.435μm signatures are present on the north pole residual cap (NPRC) and surrounding bright outliers, but not the dark dunes that surround the NPRC, since these will be warmer during this period. A particularly thin covering of CO2 ice can be seen on the dunes at 81°N, 122°E from Ls = 11°–23°.
 The width of the H2O ice annulus (Figure 3b, third column) decreases markedly in this period; by Ls = 4°–11° it has decreased to around 2° of latitude in width.
4.1.4. Asymmetric Retraction Phase (Ls = 25°–62°)
Figures 3c and 3d cover this phase. During this phase, CO2 seasonal cap begins an “apparent” asymmetric retreat. We use the term apparent because temperatures measured by TES indicate no asymmetry around the cap [Kieffer and Titus, 2001], suggesting that the cause of the apparent asymmetric retreat is an asymmetric deposition of H2O ice onto the CO2 cap, thus obscuring its spectral signature.
Figure 8 shows a simplified diagram of the suggested explanation for the asymmetric water ice distribution [Brown et al., 2011; Mellem et al., 2012]. This aeolian process releases more water in the longitude range 90–210°E, and is potentially due to exposure of rough residual ice outliers (the outliers are termed “Mrs. Chippy's Ring” by Calvin and Titus ).
Figure 8 shows the four stages that would lead to exposure of the water ice of the rough water ice outliers. In the first stage, a midwinter scene shows CO2 ice entirely covering the deposits. In the second stage, the CO2 ice has begun subliming but the rough water ice deposits are not yet exposed. In the third and crucial stage, the water ice outliers are exposed, water ice grains are then blown on top of the CO2 ice, covering up the signature of CO2 ice around them, and creating the asymmetric distribution of water ice that is observed. In the final stage, showing a summer scene, the water ice deposits are likely to be subliming and losing mass continuously. It should be emphasized that this process is subtly (but manifestly) different from the Houben process, which hypothesized sublimation of the water ice annulus and cold trapping of gas molecules as the transport mechanism. Houben et al.  describe the Houben effect in this manner (p. 9077) “the cap edge is also the location of intense baroclinic wave storm activity [Leovy, 1973; Barnes, 1981]. The warm phases of these eddies (which pick up most of the water vapor) are associated with poleward winds. They therefore transport the bulk of the water to higher latitudes. Our computations show that this newly released moisture subsequently reprecipitates from the cold phase of the baroclinic waves and accumulates at higher latitudes.”
 However, for the model in Figure 8, the water is still in thermal contact with CO2 ice and will not sublime, therefore the deposition in this model is necessarily driven by the wind as an aeolian process. The water ice that asymmetrically covers the CO2ice in our model does not travel in the vapor phase. This makes this process different from the Houben-inspired models described byBass and Paige , Schmitt et al.  and Appéré et al. .
 The H2O ice seasonal cap continues to shrink symmetrically around the pole during this time, retreating from ∼58°N to just inside the 70°N circle by Ls = 62°.
126.96.36.199. Late Increase in CO2 Signature (LICS) Events
Langevin et al.  and Appéré et al.  reported OMEGA observations of “Late Increases in CO2 ice Signatures” (LICS) around Ls = 60°–80° and CRISM data shown here corroborates that finding. For example, Figure 3d shows CO2 ice signatures in the middle of H2O ice signatures around 75°N, 340°E and 77°N, 80°E at Ls = 50°–68°. Appéré et al. attributed these CO2 ice signatures to exposures of underlying CO2 ice as H2O ice that had been obscuring the CO2 was removed, perhaps by katabatic winds [Spiga et al., 2011]. The conditions under which these processes occur clearly require further laboratory and numerical investigation.
4.1.5. Stable Phase (Ls = 62°–92°)
Figure 3e covers this phase. Since the apparent CO2 ice signatures have disappeared at this time, we show the H2O ice index images in the second column of this figure. These show the distribution of H2O ice is inhomogenous, and just as for the CO2 ice cap, the weakest H2O index regions are over the relatively warm, dark dune regions surrounding the NPRC. This is outstandingly illustrated by the CRISM Ls = 74°–80° images which show stable H2O ice south of the dunes in the 0–90°E radial, and no H2O ice on the dunes poleward of this region. MARCI images show Shackleton's Grooves (80°E, 90°E) have been covered by dust at Ls = 80°–86° [Cantor et al., 2010], and CRISM H2O ice signatures are accordingly weakened in that image.
 On the fourth column of Figure 3e, the 1.25 μm band absorption depth is plotted. The interesting observation that becomes apparent from these 1.25 μm band maps is that the largest band depths are on the periphery of the cap and they grow throughout the Stable phase. This was not apparent from the H2O index alone, for reasons discussed earlier. The largest grain sizes correspond to Korolev and the water ice outliers (Mrs. Chippy's Ring), which strengthens the argument that the spectra in Figure 4c are revealing H2O ice metamorphism which is more rapid in the equatorward (warmer) ice deposits (e.g., those at 150°E, 75°N).
4.1.6. Comparison to TES Observations of MY 24–25 Recession
Kieffer and Titus  showed that the thermal infrared coverage by TES of the springtime recession was symmetric and that stable CO2 ice temperatures completely disappeared from the cap at Ls = 78°. This slower, symmetric retreat contrasts with the asymmetric apparent CO2ice cap retreat observed by CRISM. This supports the suggestion that cold-trapped H2O ice is obscuring the edge of the CO2 ice cap [Houben et al., 1997; Bass and Paige, 2000; Schmitt et al., 2006; Wagstaff et al., 2008; Appéré et al., 2011]. We have shown here that this process is asymmetric after Ls = 25°, although the physical process responsible of this asymmetry is not well constrained and we have suggested only a simple model of this process here (Figure 8). The nature of the daily wind regimes and H2O condensation processes clearly requires further laboratory and numerical study.
4.1.7. Comparison to OMEGA Observations of MY 27–28 Recession
Appéré et al.  observed the retreat of the cap using OMEGA data, and as has been reported earlier, revealed the presence of LICS events in late spring. Appéré et al. noted some asymmetry in the early spring period in the 210–270°E (compared to 330–030°E) radial (their Figure 15) and suggested albedo, H2O ice and CO2 ice all retreated faster along the 210–270°E radial. They suggested this may be due to topographic differences or weather patterns [Hollingsworth et al., 1996]. Our results suggest the asymmetry is only in CO2 ice and possibly albedo (see Figure 3c, last row) and the fast CO2 ice retreat is from 270 to 30°E. This leads us to conclude that exposed rough ice in the ice cap outliers may be transporting material to other regions of the cap (Figure 8).
Appéré et al.  also observed that the H2O ice annulus extent was relatively large at in the middle-to late winter (they observed it was 6° wide at Ls = 330°) and decreased in width (they observed it was 2° wide at Ls = 350°). This roughly corresponds to CRISM observations of the span of the H2O ice annulus prior to Ls = 25° reported here. Some differences between the OMEGA study and the CRISM data reported here may be due to interannual variability.
4.2. Interpretation of H2O Grain Size Results
 Histograms showing the water ice index are shown in Figures 7a–7d. A histogram of the 1.25 μm band depth is shown in Figure 7e. The histograms are divided into four different phases as discussed below.
4.2.1. Presublimation Phase H2O Ice Grain Sizes
 From Ls = 304°–336°, midwinter of MY 28, the range of H2O ice index is relatively restricted (Figure 7a). During this period, the largest H2O ice index of 0.35 was achieved at Ls = 320°–328°, however it is closely matched by the Ls = 328°–336° observations. The smallest peak modes of around 0.2 are for the earliest periods, Ls = 304°–320°. Assuming they are due to H2O and dust alone and our models (Figure 2a) are correct, this would correspond to minimum apparent grain sizes slightly less than 10 ± 1 microns.
4.2.2. Early Spring Phase H2O Ice Grain Sizes
 As spring begins to warm up the north polar region (from Ls = 3°–25°, MY 28) the H2O index increases markedly, quickly adopting values roughly from 0.2 to 0.5, with modes around 0.35–0.42. Interpreting these values using Figure 2a indicates they correspond to a grain size range of 10–40 microns, with modes of 30–35 microns.
 During Ls = 25°–62°, MY 30, H2O index profiles for polar ice continue to increase. The maximum values exceed 0.6 and minimums move up to around 0.3 (corresponding to a grain size maximum of 60 ± 6 microns and minimum of 20 ± 2 microns). The modes for the last four periods (Ls = 37°–62°) are tightly bound around 0.5 (corresponding to grain sizes of 40 ± 4 microns).
4.2.4. Stable Phase H2O Ice Grain Sizes
 During the late spring stable period, the relative band depths are also the most stable and constricted. The H2O indexes have all increased above the values in the asymmetric period. Practically all of the H2O ice index values are between 0.7 (grain size of 100 ± 10 microns) and 0.4 (grain size of 30 microns). The H2O ice index mode is around 0.55 (corresponding to a grain size of 50 ± 5 microns).
 As discussed earlier, because the H2O index range of the Stable phase (late spring) exceeds 0.6, we consider it unreliable as an indicator of water ice grain size. In Figure 7e we have plotted the 1.25 μm band depth for five periods of the Stable phase. The values range from a minimum of 0.075 (corresponding to around 300 ± 30 microns from Figure 2b) to just over 0.2 (corresponding to 1200 ± 120 microns), with a mean of around 0.13 (corresponding to a grain size of 600 ± 60 microns). These grain sizes are well in excess of those estimated by the H2O index plot (Figure 2a) but we believe they are closer to reality due to the severe damping of the 1.5 μm band when the water ice grain sizes become this large.
4.2.5. Interpretations of Evolution of H2O Ice Grain Sizes
 During the Ls = 304°–92° midwinter to springtime period, H2O ice grain sizes are continuously growing. This is in contrast to the CO2 ice grain sizes in the south polar residual cap reported by Brown et al. . The increase in grain sizes might be explained in two ways: (1) by sublimation of fine grained ice and exposure of large grained H2O ice or (2) by metamorphism/sintering of H2O ice grains during this period. Both options could explain our observations, and both these processes require further laboratory and numerical study.
4.2.6. Comparison With Previous Grain Size Investigations
Kieffer  used a simple H2O ice metamorphism model to predict that if the north polar cap is undergoing net annual sublimation then late summer grain sizes would grow to 100 microns or more. Clow  predicted H2O Martian ice grains of 50 microns in size may grow to 1000 microns in size in the midlatitudes by thermal metamorphism or sintering in one summer season.
Langevin et al.  reported observations of the growth of H2O grains in the north polar cap for early summer (Ls= 93°–127°). Their study plotted spectra for three locations and used radiative transfer models to suggest sub-100 micron grains at Ls = 93° were being replaced by 700–800 micron sized grains at Ls = 127° at the Point B locality. They suggested 700 micron to 1 mm water ice grains were present at Point C in Korolev at Ls = 93.6°. Our estimates are largely similar, however we report slightly larger grains (with a maximum of 1.2 mm) because we are using the 1.25 μm band depth method devised for this study. Langevin et al. suggested thermal metamorphism could not be responsible for this growth (∼100 micron to ∼1 mm in 1 Earth month) and suggested that fine grained ice was being sublimated away during this period to reveal older, coarser grained ice beneath. CRISM northern summertime results will be the subject of future study.
Appéré et al.  reported modeling 200 micron H2O ice grain deposits on top of CO2 ice deposits in order to obscure underlying CO2 ice signatures when CO2 grains were 7 cm, however as they acknowledged, determining the grain size of CO2 ice in the presence of large amounts of water ice is a difficult challenge because the broad H2O 1.5 μm and narrow CO2 1.435 μm bands overlap significantly.
 In the south polar region, in CRISM spectra that lacked any H2O ice signatures, Brown et al.  reported CO2 ice grain sizes of up to 7 cm during southern winter. Whether large grain CO2 ice deposits are present in the north during winter remains an open question.
 We have presented the results of CRISM and MARCI mapping of the north polar midwinter and springtime during the first 3 Mars years of Mars Reconnaissance Orbiter science mapping operations.
 It has been established that the optical surface of the northern polar seasonal cap is dominated by H2O ice in much the same way CO2 ice dominates in the south [Brown et al., 2009].
 The observations reported herein have for the first time (1) mapped the springtime retreat of the seasonal cap in the north polar region over MY 28–30 and divided the retreat into Presublimation, Early Spring, Asymmetric Retreat and Stable phases, according to the distributions of the H2O ice grain size for each period (with modes of 10 ± 1, 30 ± 3, 40 ± 4 and 50 ± 5 microns, respectively); (2) established that the apparent disappearance rate of CO2 ice signatures increases rapidly in early spring and becomes spatially asymmetric after Ls = 25°, and apparent CO2 signatures disappear by Ls = 62°; (3) presented a model to account for the asymmetrical disappearance of CO2 ice signatures during the Ls = 25°–62° period (Figure 8); (4) discovered only one locality where pure CO2ice signatures are present in middle-to-late winter in the 75°N–55°N latitude range at Lomonosov Crater at 65°N, 350°E; regions poleward of 75°N were not observed in middle-late winter due to low-light conditions; (5) established that during late spring, 1.25μm band depths of ∼0.2 (corresponding to modeled grain sizes of 1.2 mm) are achieved in the water ice outlier regions (e.g., 150°E, 75°N); (6) established that H2O ice grain sizes increase from midwinter throughout springtime in contrast to the behavior of CO2 ice grain size in the south pole, which decreases in size throughout springtime [Brown et al., 2009]; (7) established that the increase in grain size at Point B during summer reported by Langevin et al.  actually commences around Ls = 86° and the grain size is remarkably stable at around 100 microns during most of spring from Ls = 34°–86°; and (8) established that the grain size of water ice in Korolev crater increases from the time the CO2 cap disappears at Ls = 46°, most likely due to thermal metamorphism.
 We thank Ted Roush for kindly supplying his palagonite optical constants. This work would not have been possible without the outstanding efforts of the CRISM Team at JHU APL and the staff at Malin Space Systems. NASA grant NNX08AL09G partially funded this investigation.