Martian dunes indicative of wind regime shift in line with end of ice age – Nature


Investigation and instruments

The previous rover-scale exploration by Sojourner/Ares Vallis52, Spirit/Gusev53, Opportunity/ Meridiani54,55, Curiosity/Gale56,57,58,59 and Perseverance/Jezero60,61,62, provide critical information on recent climate-related processes63,64 and can also provide links to wind directions65. surface alteration environments66 and aeolian processes67,68. China’s Tianwen-1/Zhurong rover made critical remote sensing and in situ observations of aeolian features in the Zhurong landing zone.

The high-resolution imaging camera69 (HiRIC) is onboard the Tianwen-1(TW-1) orbiter, which uses an off-axis three-mirror astigmatic optical system with a focal length of 4,640 mm. Three time delay and integration charge coupled device detectors are all set on the imaging plane to achieve the push-broom imaging. Each time delay and integration charge coupled device has an array of 6,144 panchromatic band pixels lined perpendicularly to the direction of the push-broom, with a pixel size of 8.75 × 8.75 μm. Topographic data products with 3.5 m spatial resolution were derived on the basis of HiRIC image data (0.7 m per pixel) collected during the parking-orbit period of the TW-1 orbiter70.

The navigation and terrain cameras71 (NaTeCam) are binocular stereo cameras mounted on the Zhurong Mars Rover, 1.8 m above the Martian surface, with a stereo baseline of 270 mm. NaTeCam is able to achieve 1.2 mm per pixel resolution when imaging a target at 3 m distance. NaTeCam images were processed through dark current correction, non-uniformity correction, absolute radiometric correction and colour correction. NaTeCam has a yaw angle range of ±178.5° and a pitch angle range from −70° to 90°, which enables ring-shot stereo imaging of the terrain surrounding the rover for topographic reconstruction.

The multispectral camera72 (MSCam) is mounted on the rover’s mast. MSCam has a focal length of 37.5 mm, with 24° diagonal field of view and 0.15 mrad angular resolution. The MSCam uses a complementary metal-oxide-semiconductor imaging detector to produce a 2,048 × 2,048 pixel image. Owing to the constraints of relay data transmission, the full image is split into 8 × 8 small windows and every subimage is composed of 256 × 256 pixels. In the default operation mode, only several windows (3 × 2 windows or 2 × 3 windows) of most interest within the full image will be uncompressed and downlinked. The 2,048 × 2,048 full image will only be downlinked in the compressed mode. Its multispectral channels are produced through a filter wheel composed of eight narrowband spectral filters (480, 525, 650, 700, 800, 900, 950 and 1,000 nm) and a broadband solar observation filter (480–600 nm). MSCam images were processed through dark current removal, flat field correction and radiometric calibration to convert observed digital number into physical radiance, and then derived reflectance from the radiance using onboard reference data of the MSCam calibration target73. The red, green and blue colours were calculated by the colour fit method74 using the eight narrowband spectral image data of MSCam.

The MarSCoDe75, which consists of a LIBS, a SWIR spectrometer and a micro-imager, is applied to probe targets on the Martian surface. MarSCoDe provides functions such as Martian surface elemental composition analysis and mineral identification through collecting active LIBS spectra (240–850 nm) and passive SWIR spectra (850–2,400 nm).

The MCS76 mounted on the rover can measure local temperature, pressure, wind and sound on the Martian surface.

The High-Resolution Imaging Science Experiment (HiRISE) mounted on Mars Reconnaissance Orbiter provides 0.25 to 1.3 m spatial resolution images77,78,79,80. Here, we used a HiRISE image (roughly 0.25 m per pixel) that covered the Zhurong landing area for dune analysis (Hires_ESP_069665_2055_RED.JP2).

Dune morphological features from remote sensing images

Using orbital images from HiRISE and HiRIC, the boundaries of 2,262 dunes were extracted from the plains around the landing site of the Zhurong rover, which covered 7 km from north to south and 5 km from west to east, after excluding disturbed regions such as impact craters, troughs and so on (Extended Data Fig. 1a). These dunes in the landing zone are sparsely distributed, disconnected and isolated, accounting for only 1.4% of the total area of the region. These dunes around the landing site are mainly represented by a ‘bright barchan-like’ shape, formed by a unidirectional wind perpendicular to the crestline. Statistics of dune morphology were extracted from their boundaries, including area of the dune boundary enclosure, the crest-ridge width (W), the ‘down wind’ length (the dimension following the inferred wind direction) (L), crestline length (rL), sinuosity (W/rL), crest to crest wavelength or spacing (λ) and wind direction indicated by dunes (Extended Data Fig. 1b). In this paper, following the geographic convention, our wind directions were given in azimuths towards which the wind blows. The azimuth of the wind direction is 0° from the north increasing clockwise to 360°.

The statistical results showed that the wind directions indicated by 95% of the bright barchans are mainly distributed in the range between 341° and 27°. The predominant wind direction was from the north-east with an azimuth of 12.5°, as illustrated in Fig. 1c and Extended Data Fig. 1c. The areas of bright barchans (95%) are between 24 and 699 m2 and the most common dune area is roughly 100 m2. The frequency decreases exponentially with increasing area. The crest-ridge width (W) of 95% of the dunes is in the 11–72 m range, crestline length (rL) also between 11 and 72 m and sinuosity between 0.8 and 1.0, increasing exponentially to 1. The down wind length (L) ranges from 2 to 13 m, and the crest to crest wavelength (λ) is around 149 m (Extended Data Fig. 1d–g). On the basis of these basic measurements, we defined plan view aspect ratio a, where a = W/L, and inter-bedform spacing s, where s = λ/L (ref. 14). In general32, TARs have a high aspect ratio (a > 6) and sometimes as high as 15. However, the a value of those in the Zhurong landing zone is slightly smaller, only about 5–6, which corresponds to small bright barchans. TARs with s = 1 were referred to as saturated, s < 2 as being closely spaced and those with higher s values as being discontinuous or widely spaced. The bright barchans in the Zhurong landing zone have higher values (s > 11), indicating that they are discontinuous or widely spaced.

Among the 2,262 bright barchans, 1,096 (roughly 48%) modified dunes can be identified from HiRIC and HiRISE images. Longitudinal dunes are developed on their western horns and a tip tail formed on their eastern horns. Among these, 500 significantly modified dunes (Extended Data Fig. 2a–c) were selected for detailed analysis. The orientations of longitudinal dunes and tip tails on the western and eastern horns of these selected dunes were analysed. The minimum, maximum and average orientations of the longitudinal dunes are 290.4°, 318.4° and 301.4°, respectively. The minimum, maximum and average orientations of the tip tails are 282.6°, 315.6° and 300.8°, respectively (Supplementary Table 1). The two orientations are similar, indicating that the eastern and western horns of the bright barchans were modified under a wind field with azimuth of roughly 300°. The west horns against the wind direction were shortened and thickened by wind denudation. The trend of eastern horns, however, was aligned with the wind direction, and therefore the horns were forced to stretch out and formed an eyebrow-like tip tail along the wind direction. Under the influence of wind, an east–west asymmetric morphology of the dune was formed.

Dune morphological features observed by Zhurong rover

Five bright barchans (dunes 1–5) were investigated in situ (Fig. 1). By using ContextCapture Master and Photoscan software, the three-dimensional models (in .obj format), digital elevation models and digital orthophoto maps of these five dunes were processed on the basis of the stereo images captured by NaTeCam.

The morphometric parameters of these dunes have been measured using 3d Max. The areas of bright barchans are in the 131–435 m2 range. The crestline length (rL) range is 29.0–66.1 m, the down wind length (L) 8.0–11.7 m. with heights of 0.5–1.0 m, representing individual examples of medium to large dunes around the landing site. The north slopes and south slopes of the five dunes range from 7.9 to 11.9° and 9.5 to 13.9°, respectively, being far lower than the common angle of repose around 30°–35°, which also indicates that these dunes have been modified (Extended Data Fig. 3a and Supplementary Table 2).

Previous studies reported on the morphology of TARs with high albedo, documenting an average crestline length of 215 m, height in the 1.5–5.7 m range and wavelengths in the 38–40 m range34,81,82. The bright barchans around the Zhurong landing site are relatively short, small and show a sparse distribution in comparison.

The digital orthophoto maps and three-dimensional models of dunes 2 and 5 were measured using ArcMap and 3d Max. The morphometric parameters of five and three longitudinal dunes, respectively, located in their west horns were obtained. These longitudinal dunes were all arranged roughly in parallel. Their trends were oriented between 278.7° and 295.9° for dune 2 and 282.4° and 294.7° for dune 5. The length of longitudinal dunes was 4.8–10.0 m long for dune 2 and about 4.0 m for dune 5. Their spacings varied from 1.4 to 3.0 m for dune 2 and 0.7–1.3 m for dune 5. The slope of the south side and north side of each longitudinal dune was not symmetric as shown by the topographic profiles (Extended Data Figs. 3b,c and 4a,b). There was some variation in the orientations of these longitudinal dunes, indicating that the wind field was variable when these longitudinal dunes were formed.

Some transverse ripples accumulated on interdune depression of longitudinal dunes due to the weakened wind strength in the down wind direction. Their orientations are perpendicular to the longitudinal dunes and have spacings of 0.3–1.0 m for dune 2 (Extended Data Fig. 3d,e) and 0.3–0.6 m for dune 5 (Extended Data Fig. 4a).

Surface properties of bright and dark sands

Several observations were made on the bright and dark sand regions on barchan surfaces by MarSCoDe and MSCam as denoted in Fig. 2a–n and Extended Data Fig. 5c,d.

Images from MSCam showed that the dark sand surfaces contained abundant agglomerated particles (roughly 3–15 mm in diameter), resembling a rough surface (Fig. 2d,i,l–n), whereas the bright sand surfaces were shown to have relatively smoothed crusts (Fig. 2b,d–f) with cracks (Fig. 2g,h,j,k). Most cracks were polygons with 3–7 sides with a length range from 0.3 to 13.5 cm (average 4.8 cm) and inner-angles ranging from 20 to 190° (average 110°). In addition, some parallel cracks developed on the denudation surfaces of the bright barchans, with orientations paralleling both the contour line and that of the longitudinal dunes. Dark sands filled in the grooves of the parallel cracks. The spacing between parallel cracks was measured at eight regions (Extended Data Fig. 3b,f), and ranges from 2 to 4 cm.

The images from MarSCoDe micro-imager showed that both bright and dark sand surfaces were shattered into a hole-like crater with a diameter of roughly 4 mm and fine powder with 100–300 μm in diameter after being hit using LIBS laser shots (Fig. 2f,n and Extended Data Fig. 5c,d), indicating that they represented cementation of small sand grains formed under some unknown mechanism.

The roughness of dune surface was analysed by computing the dissimilarity property of the grey level co-occurrence matrix (GLCM)83 derived from dune surface images. In total, ten image regions of dune surfaces were clipped from MSCam data and each image region was 200 × 200 pixels in size and had a uniform spatial resolution of 300 μm per pixel after resampling. The dissimilarity of three bright sand image regions ranged from 13.8 to 21.6 and averaged 17.0, whereas the dissimilarity of seven dark sand image regions ranged from 25.9 to 38.2 and averaged 30.4, nearly twice greater compared to bright sand; this indicates a significantly rougher surface on the dark sand surface.

At adjacent locations, the average value of reflectance increases in bright sand relative to dark sand in eight bands of MSCam and is greater than 15% (Extended Data Fig. 6).

Compositions of bright and dark sand surfaces

Reflectance spectra of dunes 1–3, from 850 to 2,400 nm, were derived from corresponding SWIR radiance data through wavelength and absolute reflectance calibrations using an onboard calibration target (a white board) and ground laboratory data84. The reflectance spectra were further smoothed by a sliding 15-pixel average filter and compared to typical Martian mineral spectrum recorded by standard mineral spectra in the RELAB spectrum library85.

The reflectance spectra were normalized to 1.0 reflectance at 2,300 nm. The spectral features of bright and dark sand both showed strong absorption features around 1.95 and 2.22 μm (Extended Data Fig. 7a). The parabola-fitting results for 1.95 and 2.22 μm absorption-band positions showed that the spectral features are similar to that of hydroxylated Fe-sulfate (Fe(OH) SO4), gypsum (CaSO4·2H2O), polyhydrated sulfate (MgSO4·nH2O) and chloride ((K, Na, Mg)Cl) (Extended Data Fig. 7b,c), but significantly differed from that of carbonates or phyllosilicates (Extended Data Fig. 7d–h). These observations indicate the existence of salts that are dominated by hydrated sulfates and chlorides in the compositions of sand particles.

A natural gradient boosting probabilistic prediction model (NGBoost)86 was established using MarSCoDe laboratory data, which had been used to quantitatively predict the main element compositions of bright and dark sands. The prediction root mean squared error (wt%) for the results of seven calibration targets was 0.11–3.49%. Main oxide concentrations of bright sand and dark sand are shown in Supplementary Table 3. The predicted total main oxides sum turned out to be below 100%. Among them, bright sand is 85.78 ± 3.80% and dark sand is 90.59 ± 3.41%. The undetectable part may be related to S, H, Cl and so on. In the LIBS spectrum, the H line at 656.468 nm could be clearly recognized. S and Cl signals in the LIBS spectrum were not detected, largely due to its weak emission below the noise level of MarSCoDe.

The composition of bright and dark sands was similar to that of average Martian soil87, close to basalt composition. The bright sands contain more Mg and more undetectable parts than the dark sands. Therefore, the bright sands may have more hydrated sulfate or chlorides than the dark sands.

Estimating AMA of the bright barchan dunes

The CSFD method was used to acquire estimated AMAs of the bright barchans within the Zhurong landing site. On the basis of the derived 2,262 dunes, the CraterTools88 on the ArcGIS platform was used to correctly extract craters superimposed on the dunes (Extended Data Fig. 8b,c). During the mapping of craters, we were very careful to avoid obvious crater chains and clusters especially near large primary craters, to exclude secondary craters. We undertook a comprehensive search for any primary craters that might have polluted our count areas with secondary craters and found no supporting evidence for this potential effect. We excluded craters mantled by the dunes (Extended Data Fig. 8a). Through cross-checking by several experts, we more cautiously eliminated a few impact craters (Extended Data Fig. 8e) that may easily be misjudged.

For the age dating of Martian surface based on CSFD, the chronology function model89 and the impact crater production function model90 is offered, showing our current best-estimate production size-frequency distribution curves down to D = 1 m. With the improvement of image resolution, metre-scale diameters craters were used for dune dating40,91. In this study, 38 effective craters (with diameters of 1.6–8.6 m) were identified (Supplementary Table 5). Considering that craters degrade more quickly on the surface of Martian barchan dunes, the dating error may be larger than that of rock surfaces. Instead of fitting a dating curve to give a certain model age value with error (±uncertainty) (Extended Data Fig. 8g), all age data values and their error bars are constrained between fitting upper and lower isochrons (Extended Data Fig. 8h) to help constrain the age of the dunes. All age data values and their error bars of bright barchans within the Zhurong landing site range from roughly 0.4 to 2 Myr (Extended Data Fig. 8d and Supplementary Table 4). To further constrain the rationality of the CSFD dating results, we also used the crater retention age method mentioned in Reiss et al.39 to estimate the retention age of the largest crater (roughly 8.6 m in diameter) on dunes, acquiring a roughly 1.7 Ma as rough upper limit of the dune stabilization age, which is roughly consistent with the results obtained by CSFD.

Warner et al.37 have clearly shown that interpreting AMAs from impact CSFD data is made much more robust by (1) larger surface count areas and (2) larger numbers of counted craters, and the clear exclusion of secondary craters in the counts. With these caveats concerning the small area available for analysis37, we tentatively interpret the time of bright barchan activity to have ceased in line with end of ice age, and the uncertainties emphasize the need for many Mars sample return missions to increase the precision of an age estimate on Mars.

With the Zhurong landing area as the centre and extending roughly 2–3° each to the north and south, four flat areas were selected for dune dating (Extended Data Fig. 8i). Areas characterized by obvious undulating terrain (for example, large impact craters and troughs), which may influence the local winds, were avoided. Supplementary Table 4 shows the dating information of these areas, which vary from 7 to 19 km2 in areal extent, with the number of dunes ranging from 26 to 1,027. Northwards from the Zhurong landing zone, the dunes in N1 and N2 are less widely distributed, and the number of superimposed impact craters is very few (only one or even none in Supplementary Table 4). It is therefore difficult to date dune activity in N1 and N2. Southwards from the Zhurong landing zone, the number of superimposed impact craters in region S1 is eight (Extended Data Fig. 8f and Supplementary Table 6) and in region S2 is two. Noting the caveats described above, the CSFD curves (Extended Data Fig. 8d) tentatively indicate that the cessation time of bright barchans activity is similar to that of bright barchans in the Zhurong landing zone.

Surface temperature and water vapour partial pressure at the Zhurong landing site

The surface temperature and water vapour partial pressure of the average solar scenario at the Zhurong landing site were simulated by using a GCM35,36. The simulated results are 182–279 K and 0.04–0.12 Pa (Extended Data Fig. 9a,b), respectively. The Modified Tetens equation (1)92, which described the relationship between temperature and saturated water vapour pressure, was used to calculate the water frost temperature.

$$E=6.112\times {{\rm{e}}}^{17.67\times \frac{t}{t+243.5}}$$


where t = T - 273.15 in °C, T is absolute temperature in Kelvin, and E is the saturated water vapour partial pressure in hPa. Let E = 0.0004 hPa or E = 0.0012 hPa, and substitute into equation (1), to obtain the t = −86 °C (187 K) or −79 °C (194 K). Because saturated water vapour pressure should be greater than or equal to water vapour partial pressure, the water frost temperature of the Zhurong landing site must be greater than or equal to 187–194 K according to equation (1).

A global map of H2O frost temperature was calculated using seasonally averaged Thermal Emission Spectrometer (TES) water vapour column abundance. According to the map, water frost would be present at low latitudes of Mars, and the water frost temperature at the Zhurong landing site93 was 197–199 K.

For the frost mixed with salt, when the temperature exceeds the eutectic temperature of the mixture, the frost melts and forms liquid saline water. Assuming the soil of the Zhurong landing site contains a mixture of MgCl2 and MgSO4, the eutectic temperature94 would be 238 K. According to the GCM simulation, the surface temperature of the Zhurong landing site could cover a range of 182–279 K. The simulation results were in good agreement with the measurement results captured by the MCS onboard the Zhurong rover (Extended Data Fig. 9c,d).

Compared with the water frost point and eutectic temperature of liquid brine, the Zhurong landing zone should be characterized by conditions in a Ls season 225 and 240, in which the water vapour could have been emplaced as frost and the frost could also melt into liquid brine.

Surface pressure, wind directions and magnitudes at the Zhurong landing site

The surface pressure, wind directions and magnitudes of the average solar flux at the Zhurong landing site were simulated by using a GCM35,36. During the in situ investigation, MCS onboard the Zhurong rover worked continuously for 5–105 min near noon on Mars, and measured local surface pressure and wind field.

The average from the daily MCS measurement data agrees with the simulated daily variation of wind directions, wind magnitudes and pressure (for example, Sol14/Ls51.0, Sol92/Ls86.1 and Sol1223/Ls148.9) (Extended Data Fig. 10e,g,i).

From Extended Data Fig. 10d,f,h, the GCM simulated results and MCS measured data in terms of seasonal changes were also consistent.

In addition, the wind speed conditions to form dunes should be at least 15.07–24.09 m s−1 (ref. 95). From the GCM simulations in Extended Data Fig. 10d,f, it is clear that during Ls225°–270°, the wind direction is the north-east wind (azimuth 30–90°) and the wind speed meets the above-mentioned threshold, which is favourable for barchan dunes deposition. During recent ice ages at higher obliquity cases, the north-east winds would be stronger and more favourable for the formation of barchan due to the deep Hadley circulation, whereas the dark longitudinal dunes were formed under north-west winds. Although the GCM simulations showed the presence of north-west winds at the Zhurong landing site, their simulated wind speeds did not reach the above-mentioned threshold. However, there were five dust storms that passed through the Zhurong landing site, which travelled towards the south-east, during the 1999–2014 period51. These dust storms from the north-west were favourable for the modification of barchan dunes and the development of dark longitudinal dunes.

Source link