Experiments On Sublimating Carbon Dioxide Ice And Implications For Contemporary Surface Processes On Mars

Carbon dioxide is Mars’ primary atmospheric constituent and is an active driver of Martian surface evolution. CO2 ice sublimation mechanisms have been proposed for a host of features that form in the contemporary Martian climate. However, there has been very little experimental work or quantitative modelling to test the validity of these hypotheses. Here we present the results of the first laboratory experiments undertaken to investigate if the interaction between sublimating CO2 ice blocks and a warm, porous, mobile regolith can generate features similar in morphology to those forming on Martian dunes today. We find that CO2 sublimation can mobilise grains to form (i) pits and (ii) furrows. We have documented new detached pits at the termini of linear gullies on Martian dunes. Based on their geomorphic similarity to the features observed in our laboratory experiments, and on scaling arguments, we propose a new hypothesis that detached pits are formed by the impact of granular jets generated by sublimating CO2. We also study the erosion patterns formed underneath a sublimating block of CO2 ice and demonstrate that these resemble furrow patterns on Mars, suggesting similar formation mechanisms.


25
There are several factors which contribute to uncertainties in our DEM measurements. Firstly there are errors due to imprecision 26 in construction and measurement of our scale objects (calibration targets). The construction of the objects was much more 27 accurate than the measurement of their locations which we estimate as better than ±1 mm. Additionally, changes in relative 28 humidity in the laboratory environment, though small, may have caused the wooden triangle to expand and contract slightly 29 during imaging and external vibrational disturbances cannot be entirely accounted for. We evaluated these errors in the SfM 30 model using a bootstrap approach. We randomly perturbed the measurements of the calibration targets by ±1 mm and then 31 re-ran the SfM software. As one would expect, if enough images are used and the targets well located, the outputs of the SfM 32 varied by no more than ±1 mm in each direction. 33 Although care was taken to sample the flat bed far enough away from the pit site for grain transport to affect the initial 34 surface measurement when measuring pit depths, sediment transport during venting may have introduced a small uncertainty 35 due to potential vertical accretion of ∼2 grains in each case where a block was placed onto the surface and in the cases where 36 the block was slid onto the 75-150 µm and 160-212 µm grains. An upper estimate of 10 grains may have accreted onto the 37 flat surface when a block was slid onto 4-45 µm and 45-90 µm granular bed. However, in all cases these fall within the error 38 estimated for our measurements.

39
of the high albedo grains to lighting differences and reflection from the container in some cases, may have led to noise affecting 41 our measurements, particularly in the flat regions used to measure pit depth. We estimated the contribution of noise to our 42 uncertainties in each case by drawing 5 topographic profiles of ∼10 cm very closely together across a flat region of interest. 43 We detrended and averaged this data and estimated noise by calculating the standard deviation in each case. The uncertainty 44 provided by noise is included in Table 1. Additionally, despite zooming to one pixel to take our vertical and horizontal 45 measurements, operator error may have occurred. Having tested our approach by making repeat measurements, we estimate the 46 contribution of this error to be ±1 pixel in each case.
For a pilot run of the first trial of experiments, we included an additional scale which was established by placing a plastic 48 ruler with millimetre graduations on the inner wall of the container which was visible in multiple overlapping images. This were the greatest errors involved in our measurements and we have used these as an approximation of z error within our models.

64
These data are included in Table 1.

66
A complete discussion of the scaling is beyond the scope of this paper and requires a detailed mathematical model. Such an analysis for the levitation of blocks has been presented previously 1 . For the geomorphological effects we build on this work and apply dimensional analysis. The key observation is that behaviour is extremely sensitive to grain size. On Earth and even on Mars, grains with radius r < 1 mm are roughly in the Stokes regime and have fall velocity where ρ s is the density of the particles, ν the dynamic viscosity of the atmosphere and g gravity. ρ s and ν are roughly the same on Earth and Mars despite the huge difference in atmospheric density. To compensate for the difference in gravity the particles need only be reduced in size by the ratio g T /g M = 9.8/3.7 = 0.61. The thermal response is given by 1 where T is the initial temperature of the sand, T s is the sublimation temperature, t is time and I is the thermal inertia. On Mars shown that the other differences should not be significant if the grain size is reduced to balance gravity. Vertical accuracy was determined using the following formula 4 : Where EP is the expected vertical precision, ∆p is the sub -pixel matching quality taken as the total RMS found in the  For the calculations of uncertainty on terminal pit area expressed as error bars in Figure 5a, we simply propagated horizontal uncertainty as follows: where σ A is the uncertainty on the area of each pit, r i is the radius of the primary terminal pit (or additional radii of secondary 112 and tertiary terminal pits) and δ H is the horizontal uncertainty on the DTM in each case. Since we used an estimate of 0.2 for ∆p, the estimate of vertical precision is possibly on the low side. We thus take the 125 estimated vertical precision for this DTM to be better than 50 cm.  Table 3. Model parameters when a ruler was used as an additional reference scale. Because manually placing markers was deemed a less accurate approach, the auxiliary scale was not used for the models reported in this manuscript.