Analysis of seasonal data from the Mars Global Surveyor Thermal Emission Spectrometer (TES) shows a significant increase in thermal inertia during autumn, winter and spring in the middle and high latitudes of Mars. At each stage of the northern seasonal polar cap's recession in spring a distinct high thermal inertia (HTI) annulus arises around the cap's edge. Within this annulus, we estimated and mapped the springtime water ice content in the daily thermal skin depth layer using spring and summer values of the thermal inertia in TES surface footprints. The results show that the average water content in the surface soil within the HTI annulus varies from ∼5 vol % at the early stages of the seasonal polar cap retreating (Ls = 340°–360°) to ∼1 vol % at later stages (Ls = 60°–70°). Maximum values of water ice content within the HTI annulus occur at Ls = 0°–20° (2–6 vol %) and Ls = 20°–40° (4–10 vol %). We analyzed the temporal and spatial relationship between the HTI annuli and the water ice (WI) annuli at the edge of the northern seasonal polar cap. The water ice within the WI annuli was mapped using a water ice spectral index (the absorption band depth at the 1.5 μm wavelength) derived from the OMEGA (Observatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité) imaging spectrometer aboard the Mars Express spacecraft. Recent OMEGA observations show that the WI annuli formation arises only around the retreating northern seasonal polar cap, never around the retreating southern seasonal cap. For this reason our study is confined only to the northern hemisphere. The observed relationship between the HTI and WI annuli in the northern hemisphere of Mars indicates a close physical interdependence between these two phenomena. Our results confirm that the seasonal permafrost exposed by the retreating northern polar cap (within the HTI annuli) is actively involved today in the condensation and sublimation processes in the modern water cycle on Mars. The water abundance in this annual condensation/sublimation cycle within the active layer may be an order of magnitude higher than all the annual water vapor abundance in the Martian atmosphere. Our results are consistent with existence of the positive mass balance of the northern permanent polar cap in the modern climatic period.
 One of the key aspects for understanding the modern water cycle on Mars is solving the question of water's role in the condensation and sublimation processes associated with changes both in the seasonal polar caps and in the surficial permafrost layer. The phenomenon of the water ice condensation within the surface layer of the Martian regolith may be related to the temporal formation of the seasonal permafrost. Based on theoretical considerations [Farmer and Doms, 1979; Schorghofer and Aharonson, 2005] it has been suggested that wintertime condensation of water ice or frost in the subsurface regolith layer (to 1 m depth) may produce a seasonal subsurface ice-bearing layer (that is seasonal permafrost) from the polar regions down to latitudes of ∼30° in both hemispheres. Seasonal permafrost on Mars to some extent represents an analog to the active layer in terrestrial permafrost areas [Washburn, 1973], which are characterized by a seasonal melting-refreezing cycle. On Mars, however, seasonal variations in the active layer are characterized by a sublimation-recondensation cycle. During spring extensive areas of the Martian seasonal permafrost must move gradually poleward as water ice sublimates, following the edge of the retreating seasonal polar caps that consist primarily of CO2 ice.
 The water ice annulus (WI annulus) formation was first suggested theoretically [Houben et al., 1997] and subsequently confirmed from TES and THEMIS observations during the recession of the northern seasonal polar caps [Kieffer and Titus, 2001; Titus, 2005; Wagstaff et al., 2008]. It was found that the extent of the visible seasonal cap is always larger than the thermally determined extent of the CO2 ice cap [Kieffer and Titus, 2001; Titus, 2005]. This difference was explained by the presence of a warmer and brighter annulus at the edge of the retreating CO2 ice cap, and water ice has been suggested as the primary constituent of this annulus [Kieffer and Titus, 2001; Titus, 2005; Wagstaff et al., 2008]. Shortly after, the occurrence of water ice in the WI annulus on the edge of the retreating northern seasonal caps was confirmed based on the near-infrared spectral observations by the OMEGA imaging spectrometer aboard the Mars Express spacecraft [Bibring et al., 2005; Schmitt et al., 2005; Langevin et al., 2005; Appéré et al., 2008]. Detailed mapping of the WI annulus evolution using OMEGA data for the Ls interval from 0.4° to 10° [Appéré et al., 2011] has shown that the WI annulus forms during the northern seasonal polar cap recession is due to the combination of three different water ice sources: the H2O frost condensed on the ground during autumn, H2O ice grains included in the CO2 ice layer during its winter condensation, and H2O vapor cold trapped over the CO2 ice layer during the spring retreat of the seasonal cap. Recent detailed mapping of the water ice spectral index (1.5 μm band) around the southern and the northern seasonal polar caps during their spring retreats [Langevin et al., 2007; Appéré et al., 2011] convincingly shows that WI annuli form only around the northern seasonal polar cap, and not around the southern seasonal cap. The WI-like annulus was observed only weakly on the edge of the southern seasonal cap in early southern winter (Ls = 115°) with a higher water ice signature close to the Hellas basin [Langevin et al., 2007]. During the late winter and spring stages of the southern seasonal cap retreat, the water ice signature was observed only as patches embedded in the CO2 ice cap. Such a difference in water content is indicative of the existing asymmetry between the retreats of the northern and the southern seasonal caps in terms of the water ice sublimation/condensation processes during the spring retreats of the polar caps. The convincing evidence that the WI annuli exists only around the northern seasonal cap was also confirmed using TES and THEMIS observations [Kieffer and Titus, 2001; Titus, 2005; Wagstaff et al., 2008].
 The presence of such asymmetry demonstrates distinctively that the processes of water ice sublimation/condensation in the northern hemisphere during the winter-spring period are more intensive and that the water abundance, involved in the WI annuli formation, is apparently more abundant than in the southern hemisphere. The most obvious reason for of the observed asymmetry between the spring retreats of the northern and southern seasonal caps is a larger supply of atmospheric water in the northern hemisphere during winter than the southern one. This is due to the existing hemispheric asymmetry of the meridional atmospheric water transfer [Haberle et al., 1993; Clancy et al., 1996; Montmessin et al., 2004; Böttger et al., 2004]. As suggested by Clancy et al. , the eccentric Martian orbit drives the north/south hemispheric asymmetries in water ice/vapor abundances through the orbitally varying water vapor saturation altitude. During the northern summer season (which occur at aphelion), water released into the atmosphere from warming the midlatitude to high-latitude regions cannot be transported into the southern hemisphere by Hadley circulation. This is because saturation and ice crystals settling remove the atmospheric water from the upper, southward traveling branch of the circulation cell [Clancy et al., 1996]. However, all water released into the southern high-latitude atmosphere during the southern summer season (which occurs at perihelion) is readily transported into the northern hemisphere in the upper branch of the Hadley circulation due to the absence of low-altitude saturation conditions [Clancy et al., 1996].
 In this work we present results of an analysis of the seasonal changes in TES thermal inertia (TI) within the area adjoining the edge of the northern seasonal polar cap, and an estimation and mapping of water ice content within the daily thermal skin depth layer (2–10 cm deep) within the HTI annuli during different stages of the seasonal cap's spring recession. We also present results comparing the seasonal evolution and spatial extent of the HTI and the WI annuli during the spring retreat of the northern seasonal polar cap. For the purposes of this study, where we compare both the TES and OMEGA data sets, which were collected in different Martian years (1998–2004 and 2006–2008, respectively), we ignore the possibility of some their interannual variability. Studies [e.g., Cantor et al., 1998] have shown that the northern seasonal polar cap's springtime retreat is repeatable year after year, with little to no interannual variation.
2. TES Observations
 To understand how TES thermal inertia is sensitive to water condensation and sublimation processes near the edge of the retreating northern seasonal polar cap, we analyzed temporal and spatial changes in TI values from late winter/early spring (Ls = 340° to 70°) as compared to summertime (from Ls = 120° to 150°) values. We used the TES observations acquired during 1998–2004 (3 Martian years) within the latitude range 30°N to 90°N at intervals of 20°Ls. The mapped TI values for each Ls range were compared with their summertime values, which have a dry (ice-free) uppermost layer. We found that a distinct annulus 3°–10° of latitude wide with sharply increased values of TI forms around the seasonal cap periphery at each stage of the cap's recession (Figure 1). It is notable that during the recession period this high thermal inertia (HTI) annulus follows the retreating seasonal polar cap edge migrating to higher latitudes as sublimation of the CO2 ice cap progresses, becoming smaller in diameter and narrower in width (Figure 1). From Ls = 340°–360° to Ls = 60°–70° the width of the HTI annulus decreases from ∼8°–10° to ∼3°. In addition, the equatorward edge of the HTI annuluses moves poleward at a rate of ∼3°–5° latitude during each 20° Ls. This corresponds to 4.5–7.5 km per Martian day. As one stage of the spring retreat changes to next the previous HTI annulus completely disappears and the thermal inertia values within the area decrease steadily as they approach summertime values. After Ls = 70° the HTI annulus completely vanishes.
 As shown recently [Kuzmin et al., 2008, 2009], the strong increase of the thermal inertia values observed during winter in the area adjacent to the edge of the seasonal polar caps is indicates an increase of water ice in the surface layer (corresponding to the daily thermal skin depth layer of 2–10 cm) due to the seasonal condensation of water ice. We suggest that the presence of an HTI annulus during the different stages of the spring represents a circular belt of the remnants from the seasonal permafrost layer, whose surface has just recently been exposed from under the retreating seasonal polar cap edge.
 At present the question of how the HTI annulus forms around the edge of the southern seasonal polar cap during its spring retreat is not resolved. As was shown in [Kuzmin et al., 2009], the presence of a noticeable HTI annulus around the southern seasonal cap was also observed during southern winter (Ls = 120°–150°). This HTI annulus was notably narrower as compared to the wintertime northern HTI annulus. The fact that a southern HTI annulus exists during southern winter points to the presence of the seasonal permafrost in the surficial regolith layer. However, it was impossible to track the formation of the southern HTI annulus during the cap's spring retreat because the Martian atmosphere was very dusty due to the dust storms activity on Mars [Smith, 2002, 2004]. The atmospheric dust made unreliable the thermal inertia values derived from the TES bolometer measurements thus limiting our ability to study the evolution of the phenomenon during spring retreat stages of the southern seasonal cap. So the question of the possibility of an HTI annulus forming around the retreating southern seasonal cap and its extent continues unsolved. Although not directly observed, an HTI annulus formation may take place during the spring retreat of the southern seasonal cap but on some smaller scale. Considering the much smaller supply of water vapor in the wintertime southern hemisphere [Haberle et al., 1993; Clancy et al., 1996; Smith, 2002, 2004], it is reasonable to suggest that the formation of an HTI annulus around the southern seasonal cap may be much weaker than around the northern seasonal cap. Thus a north-south asymmetry in an HTI annulus formation may occur, but not as obviously as in the case of WI annulus formation.
3. Mapping the Groundwater Ice Content Within the HTI Annulus
 To estimate and map the water ice content in the surface soil layer within the HTI annulus, we used the method developed recently by Kuzmin et al.  to determine the wintertime increase of water ice in the surface layer using TES TI data. The method is based on the comparing summer and winter values of thermal inertia derived from TES observations of the same geographic location. In present work, we compared the summer thermal inertia values (during Ls = 120°–150°) with those measured during different stages of the seasonal cap's springtime recession. The TES TI data were based on TES observations collected during 1998–2004 (covering 3 Martian years). These were used for spring and summer data where the location of the TES footprints differed by less than 0.03°. In this way we obtained TI pairs for spring and summer stages that were used to solve the quadratic equation discussed below. According to Schorghofer and Aharonson , the thermal inertia (I2 = kρc) of the two-component mixture (soil + ice) may be described by the parametersρc = ε ρicecice + ρdrycdry and k = ε kice + I2dry/(ρdrycdry), where the parameter ε represents the water ice volume fraction. After substitution of the parameters into formula (I2 = kρc) the quadratic equation aε2 + bε + c = 0 can be written, where a = ρicecicekice, b = ρdrycdrykice + ρicecicekdry and c = I2dry − I2s. Idry and Is represent the mapped thermal inertia values for summer and spring seasons, respectively. All parameters for water ice were calculated for T = 180 K (ρice = 927 kg m−3, cice = 1397.4 J kg−1 K−1, kice = 3.18 W m−1 K−1). For dry soil we used cdry = 836 J kg−1 K−1, and ρdry was calculated as function of Idry according to Mellon and Jakosky . In this way we calculated coefficients of the above equation for each coincident summer and spring TES pair and then calculated the value of the water ice volume fraction ε. The values of ε (in vol %) were then compiled into the map of the springtime water ice abundance in the surface layer of the Martian soil equal to daily thermal skin depth layer (2–10 cm).
 The resulting estimated water ice content of the HTI annulus is illustrated in Figure 2. As Figure 2 shows, the width of the water ice region in the surface layer gradually decreases from 8°–10° to ∼3° during the spring recession of the seasonal polar cap (from Ls = 340°–360° to Ls = 60°–70°). This observed decrease in width of the seasonal permafrost remnant may be due to the gradual increase of solar radiation during spring that accelerates the sublimation of ground ice from the active layer. The average change in water ice content within the HTI annulus during different stages of the spring retreat of the seasonal polar cap is shown in Figure 3. The average water ice content (for the values >1 vol %) in the surface soil layer within the HTI annulus decreases from ∼5 vol % from the beginning of the spring recession (the Ls = 340°–360°) to ∼2 vol % at the end (the Ls = 60°–70°). From Ls = 20°–40° a relatively higher value of the average water ice content is observed (see Figure 3). It is possible that the increased water ice content in the active layer during this stage may be due to increased condensation of the water ice within the active layer at latitudes 56°N–62°N during the preceding autumn and winter.
Figure 4 shows the frequency distribution of all the water ice content values within the HTI annuli formed as the seasonal cap recedes. As one can see, the frequency distribution of the water ice content within each of the mapped HTI annuli is differs. Whereas the frequency distribution of the water ice abundance during Ls = 340°–360° and 40°–70° is similar, the distribution varies significantly from what occurs during the Ls = 0°–40°. The prevailing values of the water ice content during the Ls = 340°–360° and Ls = 40°–70° are 0–2 vol % (frequency 47.84%), while during Ls = 0°–20° and 20°–40° the prevailing values are higher in the range 2–6 vol % (frequency 52.52%) and 4–10 vol % (frequency 66.29%), respectively. Disappearance of the HTI annulus after Ls = 70° indicates the upper several centimeters of the Martian active layer have completely lost the water ice that accumulated in the layer during the previous autumn and winter seasons. The average value of the water ice abundance involved in seasonal formation of the HTI annuli is ∼2% by volume. Since both the area of the HTI annuli seasonal formation and total water ice content within surficial layer are significant the next question arises from here: what the order of magnitude of a potential amount of the water ice content is involved in the seasonal condensation/sublimation cycle on Mars within the surficial layer? We estimated a potential amount of the water ice content taking for calculation the square value of the latitude belt of the HTI annuli formation during spring retreat period of the seasonal polar cap (50°N–70°N), the average daily thermal skin depth layer of surface soil (∼3 cm) and average value of the water ice abundance involved in the seasonal formation of the HTI annuli (2% by volume). At such constrains the estimated potential quantity of the water ice in indicated latitude belt may amount to the value about 7.5 km3, which is significantly more than total quantity of water vapor in the Martian atmosphere (∼1–2 km3) [Farmer and Doms, 1979; Titov, 2002]. If we exclude from the calculated water ice amount the total atmospheric water content, then recondensation of the remaining part of the potential water abundance within the area of the perennial northern polar cap may result in the deposition of a new annual water ice layer about ∼0.7 cm thick.
 The estimated values of the water ice content within the HTI annulus directly depends on the seasonal values of the thermal inertia, and their accuracies will be determined by the uncertainty of the derived thermal inertia, which is estimated to be <6% [Mellon et al., 2000]. This uncertainty is introduced at three major sources of uncertainty [see Mellon et al., 2000]: in the TES measured temperature (2.7%), in the thermal model used to generate the lookup table (1.4%) and in the interpolation scheme (inertia, albedo, pressure, opacity, hour, season, latitude (all together 1.9%). So the total uncertainty should be considered an upper limit estimate, since it was assumed [Mellon et al., 2000] that all errors are positive and systematic. The calculated accuracy of the ice content within the HTI annuluses (see Appendix A) varies lineally from approximate 18–20% for a water ice abundance of <3 vol % to 12% for a water ice abundance of 16 vol %.
4. Relationship Between the HTI and the WI Annuli
 To define the spatial relationship of the HTI annuli (the seasonal permafrost remnants) to the WI annuli, we mapped the surface water ice cover at the edge of the retreating northern seasonal polar cap. For this we used the spectral index of the water ice (the depth of the absorption band on the 1.5 μm wavelength), derived from the spectral observations by the OMEGA imaging spectrometer aboard the Mars Express spacecraft [Bibring et al., 2005; Langevin et al., 2005]. The CO2 ice absorption band depth at the 1.435 μm wavelength was used for mapping of the seasonal CO2 ice cover. The mapping of both indexes was conducted for the same spring stages of the seasonal polar cap's retreat as done for the analysis of the seasonal changes in TES thermal inertia values within the HTI annuli.
 Both the spectral indexes were calculated using the intervals between the center of the absorption band and the selected continuum regions of the OMEGA spectrum. We used the same approach for calculation both spectral indexes as was developed in the works of Langevin et al. , Pelkey et al. , and Appéré et al. [2008, 2011]. The water ice spectral index was calculated by
 To minimize the impact on the CO2 ice spectral index of both the water ice and atmospheric CO2 bands, we used the similar spectral index of CO2 ice, as presented by Langevin et al. . The CO2 ice spectral index was calculated as band strength by:
The results of the spectral index mapping are shown in Figure 5. While our mapping has been conducted using a time step ΔLs = 20°, the picture of the detected WI annuli evolution during the spring polar cap recession agrees well with the results of more detailed mapping of the annulus types conducted with the time step ΔLs = 0.4°–10° [Appéré et al., 2011]. As seen from Figure 5, at the beginning of the seasonal polar cap recession (Ls = 340°–360°) the width of the WI annulus is 2°–3° and widens as spring progresses. During spring (Ls = 0°–70°), a significant widening of the WI annulus is observed while the area of CO2 ice cover rapidly decreases. Beginning at Ls = 40°–60° the CO2 ice cover remains as separate patches while the water ice cover occupies most of the retreating cap surface. Following Appéré et al. , the water ice cover within the WI annulus may approach millimeters in thickness. Formation of the WI annulus is due to the combination of two processes occurring in the spring time. These are the condensation at the seasonal cap's edge of water vapor released by the sublimation of the water ice from the active layer (the seasonal permafrost exposed from under the seasonal polar cap cover), and second, the deposition of the water ice grains embedded in the sublimating CO2 ice as lag deposits on the edge of the seasonal polar cap.
Figures 6 and 7 show the distribution of the groundwater ice content within the HTI annuli and the OMEGA WI annuli for different periods of the seasonal polar cap recession. During winter (Ls = 300°–320°), both the northern edge of the TES HTI annuli and the external boundary of the OMEGA WI annuli closely align (Figure 6). At the beginning of spring the boundaries of both annuli types are adjoin only in some longitude sectors and become more separated later in spring (Figure 7).
Figure 8 compares of the WI and the HTI annuli seasonal dynamics. The general trend of the temporal evolution and spatial extent is very similar for both the HTI and the WI annuli. The retreat of the outer edges of the both annulus types during spring is similar, with a rate of 4.5–7.5 km per Martian day for the HTI annulus and to 6.2 km for the WI annulus [Appéré et al., 2011]. Between Ls 30° and 70° the distance between both types of annulus increases gradually, following a similar trend as the spatial difference between the southern limit of the CO2 ice deposits and the external edge of the WI annulus [see Appéré et al., 2011, Figure 16]. In this season the WI annulus is located on the edge of the seasonal cap [Schmitt et al., 2005; Pankine et al., 2010; Appéré et al., 2011], and its outer edge is usually detached from the internal boundary of the HTI annulus by up to several degrees of latitude (see Figure 7). It is possible that the lag of the internal edge of the HTI annulus behind the external boundary of the WI may be due to a more rapid sublimation of the thinner cover of the water ice on the seasonal cap's periphery as compared with the slower sublimation of the ground ice from the surficial soil active layer within the HTI annulus. However, a more likely explanation is that the internal boundary of the HTI annulus represents in fact the limitation of the TI mapping due to uncertainties of the thermal model (H. H. Kieffer, Thermal model for analysis of Mars infrared mapping, submitted to Journal of Geophysical Research, 2012) used to derive the TI values in the high latitudes of Mars during the winter-spring period. So it is more probable that the real northern boundary of HTI annulus area (at each stages of the seasonal polar cap recession) may extend up to the outer edge of the WI annulus, as it mostly occurs during the periods Ls = 300°–310° and Ls = 340°–360°. The similarity of the retreat rate and the observed spatial relationship of the two types of annuli support a close interdependence between their origin by condensation and sublimation processes taking place around the seasonal cap edge. We suggest that the sublimation of the seasonal groundwater ice within the HTI annulus (as exposed in the soil's active layer) serves as main source for both the WI annulus cover and the water vapor annuli occurring above the edge of the retreating seasonal cap [Pankine et al., 2010]. The location of these water vapor annuli detected from the TES data [Pankine et al., 2010] directly correlates with the position of the WI annuli.
 Our analysis of seasonal variations seen in TES thermal inertia has shown that the annuli of increased thermal inertia (HTI) form near the edge of the northern seasonal polar cap during the spring cap recession. Based on the mapping of the water ice content in the surficial soil (in the top 2–10 cm layer) within the HTI annuli we found that the average water ice content is ∼5 vol % in the early stage of the seasonal polar cap retreat (Ls = 340°–360°) and decreases to ∼1 vol % by the end of spring (Ls = 60°–70°). The maximum values of the water ice abundance are correlate with the HTI annulus which forms during Ls = 0°–20° (2–6 vol %) and Ls = 20°–40° (4–10 vol %). Within other Ls ranges the prevailing values of the water ice abundance in the HTI annulus are in the range 0–2 vol %. We suggest that the observed trend in the reduction of both the water ice abundance within the HTI annulus and its reduction in width (including its complete disappearance during the spring recession of the northern seasonal polar cap) may result from the combination of two processes. First, water ice steadily decreases within the active layer progressing from the middle to the high latitudes. Second, increasing solar radiation causes a gradual increase in the water ice sublimation during the period from Ls = 340°–360° to Ls = 70°.
 Our analysis of TES and OMEGA data demonstrates the existence of significant seasonal variations of water ice within the active soil layer and at the periphery of the northern seasonal polar cap during recession. Comparing the HTI and the WI annuli shows a close interdependence between these two annulus types. Both are characterized by similar rates of retreat of their outer edges during spring (4.5–7.5 km and 6.2 km per Martian day for the HTI annulus and the WI annulus, respectively). We suggest that during the northern seasonal polar cap's recession the sublimation of the water ice from the exposed seasonal permafrost layer within the HTI annulus area serves as the source of the water for both the observed increase in atmospheric water vapor and the formation of the WI annuli on the edge of the retreating CO2 ice cap that is well seen spectrally in the OMEGA data [Bibring et al., 2005; Schmitt et al., 2005; Langevin et al., 2007; Appéré et al., 2008, 2011]. Whereas the HTI annuli may represent the remnants of the seasonal permafrost exposed from under the seasonal cap, the WI annuli likely originate by recondensation of the water vapor (released due to the water ice sublimating from the active layer) and the deposition of the water ice grains embedded in the sublimating CO2 ice as lag deposits on the edge of the seasonal cap [Appéré et al., 2011]. According to our results therefore, the seasonal permafrost around the northern seasonal polar cap and under its bed is actively involved in condensation and sublimation processes as part of the modern water cycle on Mars.
 In addition to the observed north-south asymmetry in the formation of the WI annuli (and possibly of the HTI annuli), observations by the High Energy Neutron Detector (HEND) on board of the Mars-Odyssey [Mitrofanov et al., 2004; Litvak et al., 2006] also show a hemispherical asymmetry in hydrogen distribution within the surficial 1 m layer. Analysis of the HEND data shows that for the northern high latitudes the homogeneous model provides an appropriate agreement with observations [Litvak et al., 2006]. Data analysis shows that at high northern latitudes the subsurface ground ice lies very close to the surface and may be covered by a layer of dry soil at most a few centimeters thick. A different picture was found for southern high latitudes. The southern hemisphere data disagree with the homogeneous model and are in better agreement with a double-layered model for the regolith [Litvak et al., 2006]. This means that the subsurface layers rich in water ice are covered by a dry surficial soil layer about 10–20 cm thick. The presence of such a dry surficial layer in the southern hemisphere is confirmed also by analysis of the neutron data received by the Neutron Spectrometer (NS) on Mars Odyssey [Prettyman et al., 2004]. Following the work of Prettyman et al. , neutrons flux measurements in the southern hemisphere (>60°S) may be fitted by a double-layered model with 2% water (by weight) in the upper layer and 60 ± 10% water (by weight) in the lower layer. The presence of a thick dry layer of soil above layer rich in groundwater ice indicates more intensive sublimation in the southern hemisphere compared to northern hemisphere. Such asymmetry suggests the surficial soil layer of the ice-bearing permafrost in the southern hemisphere in modern time (in the annual time scale) undergoes a stronger loss of the water ice, while in the northern hemisphere the groundwater ice in the surficial soil is more stable. The modern deficit of atmospheric water, plus the higher spring and summer surface temperatures in the southern hemisphere (perihelion) appear to be main reasons for a thick dry soil layer covering a deeper water ice rich soil. Each of the three data sets (HEND, TES and OMEGA) relates to a different thickness of the surface soil layer (1 m, 2–10 cm and ∼10μm, respectively), all three show a north-south asymmetry in water ice sublimation/condensation processes. This means that the modern hemispheric asymmetry of the atmospheric circulation on Mars controls the character of the water ice sublimation/condensation processes, both on the surface and down to a depth of one meter.
 The mapping results show that the potential quantity of water involved in the annual condensation/sublimation cycle within the active layer (seasonal permafrost layer) may be higher than the total water vapor content in the atmosphere of Mars. Taking into the account the fact that the modern atmosphere of Mars has an asymmetric pattern of the meridional circulation [Clancy et al., 1996] that is dominated by the transport of water vapor (during northern winter) from the southern hemisphere to the northern one [Clancy et al., 1996; Richardson and Wilson, 2002; Montmessin et al., 2004], it seems possible that the significant part of the water vapor released by the sublimation of groundwater ice from the active layer is going to constitute the annual increase of the water ice deposits on the surface of the northern perennial polar cap. Thus it is possible that the annual mass balance of the modern northern perennial polar cap today is growing compared to that of the southern polar cap. The presence of the north-south asymmetry in terms of both the WI formation and the hydrogen distribution in the surface 1 m layer is well consistent with such suggestion.
Appendix A:: Estimation of the Relative Error of the Mapped Ice Content Values Within the HTI Annulus
 Defining the relative error of the seasonal variation of water ice abundance within the Martian soil active layer depends on determining the precision of the seasonal thermal inertia. We have derived the water ice abundance as ε = ε (I, Idry). The uncertainty errors of derived values for thermal inertias are estimated to be <6%. So, the maximum error of the estimated water ice abundance in surface soil can be given by
where ΔI = 6%I, ΔIdry = 6%Idry. In our work we used the linear parameterization of thermal conductivity: expressed as k = εkice + kdry. In this case the values for (ε) may come from solving the quadratic equation:
where а = ρicecicekice, b = ρdrycdrykice + ρicecicekdry,, c = I2dry − I2; kdry = I2dry/ρdrycdry. Differentiating equation (A2) with respect to variables I and Idry using implicit differentiation, we receive after simple manipulation the expressions for the partial derivatives:
The relative error of the mapped water ice abundance (ε) is Δε/ε. The dependence of the relative error values on the mapped water ice abundance are shown on the Figure A1.
 As one can see from the Figure A1, there is a noticeable trend of the relative error values decreasing with increasing of the derived water ice abundance values. Parameters, using by calculation both ε and Δε/ε, are given in Table A1.
Table A1. Parameters Used to Calculate the Relative Error of the Mapped Water Ice Amount in a Two-Component Mixture (Soil and Ice)
 Authors wish to express their gratitude to Yves Langevin and Tim Titus for their constructive review of the manuscript. We also thank to I. A. Kozhevnikovoy for helpful discussion. This study was supported by the Russian Foundation for Basic Research (project 10-2-00464-a) and partially by the Programme-22 of the Russian Academy of Science Presidium.