Radiation Environment and Doses on Mars at Oxia Planum and Mawrth Vallis: Support for Exploration at Sites With High Biosignature Preservation Potential

The first human missions on Mars will likely involve several astrobiology‐driven science operations, at sites with high biosignature preservation potential. Here, we present a study of the radiation environment induced by Galactic Cosmic Rays and Solar Energetic Particles at Oxia Planum, landing site of the European Space Agency ExoMars 2022 mission, and at two different locations in Mawrth Vallis, using the Monte Carlo GEometry ANd Tracking 4‐based code dMEREM (detailed Martian Energetic Radiation Environment Model). The radiation environment for solar minimum in 2009 and a period close to solar maximum during the declining phase of solar cycle 23 appears similar at the different sites, with the deepest Mawrth Vallis location having a slightly enhanced γ‐ray contribution, due to a higher modulation of fast neutrons by the more water‐rich regolith. The comparison with the Dose Equivalent from an updated extrapolation of 7+ years data from the Radiation Assessment Detector (RAD) onboard the Curiosity rover highlights the importance of input modulation conditions, some drawbacks of the galactic cosmic ray model used here, and the need to include heavy ions, the three aspects affecting differently the estimations for solar maximum and minimum. The dependence of doses on surface pressure highlights a possible influence of the different dust loading at the different sites. Estimated exposure levels for a 1‐year stay and for a short stay in Arabia Terra, the latter including a October 28, 2003 event with a fluence an order of magnitude higher than the relevant September 2017 event detected by RAD, leave reasonable to large safety margins.

Oxia Planum and Mawrth Vallis, two regions near the northern boundary of Arabia Terra, offer excellent conditions for the search of biosignatures. Oxia Planum is the final selected landing site for the next European Space Agency (ESA) mission ExoMars 2020 (now ExoMars 2022), partially because of more suitable conditions for the technical constraints such as landing and traversability, and different specific locations in Mawrth Vallis have been previously considered as candidate landing sites for ExoMars 2022 (Bridges et al., 2017), the current Mars Science Laboratory (MSL; Rogers & Bandfield, 2009) and Mars 2020 mission (Gross et al., 2017). Based on near-infrared OMEGA (Observatoire pour la Minéralogie, l'Eau, les Glaces et l'Activité) data as well as CRISM (Compact Reconnaissance Imaging Spectrometer for Mars) data, the regional compositional mapping has revealed wide deposits of clays at both sites, in particular Mg/ Fe phyllosilicates (Carter et al., 2013). In particular, Oxia Planum exhibits a layered, meters-thick nearly homogeneous Fe-Mg phyllosilicates deposit and some hematite, while Mawrth Vallis mostly displays Fe-smectites with multiple oxidation states and a relevant presence of several other hydrated minerals (Poulet et al., 2020), in principle making the site rewarding in terms of both scientific outcomes and In-Situ Resource Utilization (ISRU). Clays are known to play a major role in the association of organic matter with mineral matrices and in the stabilization and preservation of biological molecules (Keil & Mayer, 2014). About hematite, some studies have reported it has a beneficial role for the preservation of microscopic signatures of life (Allen et al., 2004), while others have reported that it offers low preservation of amino acids (dos Santos et al., 2016) and induces oxidation of organics (Ertem et al., 2017).
The radiation environment at the surface of Mars is constituted by primary Galactic Cosmic Rays (GCRs), the low intensity background radiation flux modulated by the 11-year solar cycle, and by sporadic, high fluxes of Solar Energetic Particles (SEPs), as well as by secondary particles produced in the atmosphere and the regolith via ionization, spallation and fragmentation processes. The high-energy GCRs can penetrate the standard spacecraft hulls, as well as potential planetary habitats and astronauts suits, producing dangerous secondary radiation particles, effective at either directly breaking DNA strands, or producing chemically active radicals in tissues that can also induce bond breaking. Late stochastic effects from non-lethal doses, such as cataracts, cancer, damages to the central nervous and/or cardiovascular system (Cucinotta et al., 2012a(Cucinotta et al., , 2012b) are observed over time as an increased likelihood of occurrence relative to the general population. SEPs are produced by transient events on the Sun, such as flares and/or Coronal Mass Ejections (CMEs), which trigger the acceleration of intense fluxes of (mostly) protons and electrons with energies below a few hundreds MeV. The relatively rapid rise in particles fluxes during such events and the possible occurrence of such events at any time in the solar cycle (Feynman et al., 1993(Feynman et al., , 2002 have implications for planning expeditions in which the explorers would drive considerable distances from their habitat. Events can last for hours up to many days, and give rise to deterministic effects, as the Acute Radiation Syndrome (Anno et al., 1989), skin injury, the depletion of blood-forming organs (BFOs), and death (Cucinotta et al., 2015;Wu et al., 2009).
In this work, we provide a first assessment of the radiation environment and doses at Oxia Planum and Mawrth Vallis as induced by GCRs in solar minimum (January 2009, northern fall) and close to solar maximum during the declining phase of solar cycle 23 (October 2003, dusty northern winter), and by SEPs DA PIEVE ET AL.  (Desorgher, 2005) ESA's tool dMEREM (Gonçalves et al., 2010;McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012;McKenna-Lawlor, Gonçalves, Keating, Reitz, et al., 2012). The results are compared with an updated extrapolation based on 7+ years of RAD data. The contribution of downward and backscattered radiation, the contribution of different particle type, and the variation of daily dose rates with surface pressure during the two solar modulation periods are investigated. We focus in particular on the absorbed dose, the main dosimetric quantity related directly to energy deposition in a target (locally), the Ambient Dose Equivalent (ADE), an operational quantity for measurements and radiation protection, and on the Effective Dose (ED), which is the relevant quantity for stochastic effects. Results for long and short stays, the latter including the occurrence of the October 28, 2003 SEP event, are compared to the radiation exposure limits usually recommended and to the relevant September 2017 event detected by RAD.
The regolith composition of the three sites considered in this work is based on data from the Gamma Ray Spectrometer (GRS) aboard Mars Odyssey and it is reported in Table 1 (Smith et al., 2001). The regolith is modeled as a 100 m deep layer and its composition is rather different from the one used in other studies for Gale Crater (Matthiä et al., 2016), that purely considers a default basaltic composition (SiO 2 51.2%, Fe 2 O 3 9.3%, H 2 O 7.4%), resembling a general terrestrial basalt (Nockolds, 1954).
Atmospheric temperature, density and composition were extracted at the three sites from the Mars Climate Database (MCD) (Forget et al., 1999(Forget et al., , 2006, up to an atmospheric height being h = 150 km, using the default solar minimum scenario for January 2009, and the dusty solar maximum scenario and the baseline MY26 scenario for October 2003. A postprocessing software was used to combine the surface pressure data with MGS MOLA topography (Smith et al., 2001). The period of October 2003 (close to solar maximum) and January 2009 (solar minimum) correspond to northern winter and fall, respectively. Simulations are performed for local 2 a.m. and 2 p.m.
The QGSP − BIC − HP physics list is used, which uses a quark gluon string model for hadrons at high energies (>10 GeV), the binary intranuclear cascade model for nucleons at energies <10 GeV, and, for neutrons with energy <20 MeV, the G4NeutronHP model (based on the Evaluated Nuclear Data File-database). The low electromagnetic list (Livermore models) is used. The number of simulated primary particles was 10 6 per bin.

Radiological Quantities
In this work, we focus on the absorbed dose, the ADE and ED. The absorbed dose corresponds to the average energy deposited by energetic particles in a target (locally, e.g., at the point of interest) (Mertens, 2016). In biological targets, this is considered to be predominantly due to ionization energy loss: where m is the mass of the volume of interest in the target, dϵ/dx is the (unrestricted) collisional stopping power (corresponding to the electronic stopping power, e.g., the energy lost by the particle per unit path length), f(E) is the particle fluence (E being the energy of impacting particles), and ρ is the density in the volume of interest, for instance, a detector or a human organ. The unit of D is the Gray (where 1 Gy = 1 J kg −1 ). In order to estimate the absorbed dose, not automatically printed by dMEREM, a water layer of a thickness of 10 cm was added at the ground and the energy deposition in that layer is scored, similarly to previous works (Gronoff et al., 2015).
The ED is the dose received by the full body and is expressed as the sum of the equivalent doses in all tissues and organs of the body, weighted by an organ/tissue weighting factor such that: where w T is the weighting factor of tissue T and H T is the equivalent dose for that tissue , , T R D is the average absorbed dose from radiation R in tissue or organ T, and R w is the radiation weighting factor of radiation R, which varies according to particle type and energy (ICRP, 2010). The ED is measured in Sieverts (Sv).
In dMEREM, the alternate operational quantity ADE used in instrumental measurements is also implemented, as it avoids the complexities associated with phantoms or patient anatomy (Fontenot et al., 2009). The reference phantom is the International Commission on Radiation Units and Measurements (ICRU) sphere, a 30 cm-diameter sphere made of tissue equivalent material (density: 1 g/cm 3 ; mass composition of 76.2% O; 11.1% C; 10.1% H and 2.6% N). The ADE at a location of interest in a radiation field is defined as the Dose Equivalent (DE) which would be generated in an oriented and expanded radiation field at a depth of 1 cm on the radius of an ICRU Sphere, oriented so as to be opposite to the direction of the incident radiation (Casimiro et al., 2018;Osamu et al., 1999). Differently from the equivalent dose (not studied in this work), the DE is defined as a product of Q and D at a point in tissue, where D is the absorbed dose and Q the quality factor at the point of interest, a function of unrestricted linear energy transfer (unrestricted LET, denoted as L below) in water.
Practically, in dMEREM the ADE is calculated by multiplying the fluences for each particle by previously calculated fluence-to-ambient dose equivalent conversion factors (tabulated in [Pelliccioni, 2000] and obtained with the FLUktuierende KAskade or Fluctuating Cascade) and adding all the contributions from the different particles i: The ED is also calculated in a similar manner, by applying previously calculated fluence-to-Effective Dose conversion factors applied to the distribution of particle fluence as a function of energy (Pelliccioni, 2000).
It is important to underline that the ADE and the DE reported in studies on RAD data are two distinct quantities. The latter is defined as the volume of the organ or tissue T, and   Q L is the quality factor, while the ADE is calculated at a specific point in the tissue-equivalent phantom (the ICRU sphere).

Space Radiation Environment
The GCRs fluxes for protons and α-particles for solar minimum (January 2009) and close to solar maximum during the declining phase of solar cycle 23 , obtained with the ISO-15390 model (Nymmik et al., 1996), are reported in Figure 2 (top panel). The spectra were retrieved from SPENVIS (Heynderickx et al., 2004;Kruglanski et al., 2009) and are considered as isotropic. Strong solar modulation as in October 2003 reduces the contribution of the low energy portion of GCRs protons and α-particles. Heavier primary ions have been ignored, since they contribute to only about 1% of the GCRs flux (Simpson, 1983) (and even less of the SEPs flux). However, it should be underlined that high-Z particles may interact with the atmosphere and generate secondaries, which would still contribute to the surface radiation exposure (Dartnell et al., 2007;Matthiä et al., 2016;Röstel et al., 2020).
The 28 October event and the series of events that took place at the end of October 2003-beginning of November 2003 were previously studied by several authors (Chenglong et al., 2008;Crider et al., 2005;Jiggens et al., 2014;Kozarev et al., 2010;Mewaldt et al., 2005). The differential proton fluxes for the October-November 2003 SEP events from the Geostationary Operational Environmental Satellite GOES-11 and retrieved from the Solar Energetic Particle Environment Modelling (SEPEM) reference event list (Jiggens et al., 2012) are reported in Figure 2 (middle panel). The largest solar proton enhancement occurred on the 28 October following an X17.2 class flare (Jiggens et al., 2014). This was the largest SEP event of the past 20 years in the tenths of MeV energy range (Jiggens et al., 2014). In principle, low energy SEP events are relatively shielded by the Martian atmosphere. Nevertheless, the October 28, 2003 event, which occurred during the northern dusty winter and which caused the unrecoverable failure of the Martian Radiation Environment Experiment (MARIE) instrument on board Mars Odyssey, is an interesting case to consider and to compare with the recent results reported for the September 2017 event, the most relevant event observed by MSL-RAD on Mars up to now (Guo, Dumbovi, et al., 2018a;Zeitlin et al., 2018).
The magnetic connection between Earth and Mars was good for the series of SPEs during the end of October 2003-beginning November 2003. Here, near-Earth particle measurements from the GOES-11 spacecraft have been used as proxies to estimate the overall particle doses for the October 28, 2003 event, assuming the fluxes fell off as 1/r 2 (where r is the helio-radial distance). In Figure 2 Table 2 presents a comparison between the ADE and absorbed dose daily rates (averaged over day/night) for GCRs and an updated extrapolation of the DE and absorbed dose rates based on 7+ years RAD data at Gale Crater, based on a similar previous analysis that was using fewer statistics (Guo et al., 2015). The comparison to the DE is only indicative, as here the ADE is calculated; nevertheless, we proceed as in previous works (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012) where useful information was extracted by comparing the ADE with results on the DE from different works.

GCRs Induced Daily Dose Rates and Comparison With the Updated Extrapolation From 7+ Years of RAD Data
Our results for both solar minimum and close to solar maximum show very similar ADE average daily rates for Oxia Planum and Mawrth Vallis site 1, essentially coinciding within statistical uncertainty. The ADE average daily rate at Mawrth Vallis site 2, the location with the lowest elevation, is slightly lower with respect to the ADE rates at the other sites, in particular for solar minimum, due to a stronger shielding effect by the atmosphere. We note that our ADE average daily rates for the three sites are considerably lower than the updated analysis of 7+ years of RAD data for both solar maximum and solar minimum, but in fairly good agreement with the previous extrapolation based on fewer statistics for solar maximum (around ∼350 μSv/ day, not reported in Table 2) (Guo et al., 2015). The absorbed dose average daily rate obtained for a water slab of 10 cm at Oxia Planum is 230 μGy/day for solar minimum and 110 μGy/day close to solar maximum ( Note. Values from the RAD analysis are given in plastics, no correction factor has been used as plastic is close to water in terms of its response to the energetic charged particles that dominate the radiation environment on Mars . The DE from Liulin-MO measurements during ExoMars TGO cruise (Semkova et al., 2018)   2 at lower elevation. We note an overall underestimation of the absorbed doses when compared with the extrapolated absorbed doses from the RAD analysis, similarly to the case of the comparison between the ADE and the DE.
The comparison of our results with those from the updated analysis of 7+ years of RAD data is affected by different factors, among which some differences between the initial modulation conditions given in input to our calculations and those on which the RAD analysis is based, the choice of a specific GCRs model and the neglect of heavier ions in our calculations. First, the GCRs model used by dMEREM (ISO 15390)  For what concerns the specific GCRs model, the ISO 15390 model used here has been reported to exhibit its own discrepancies with the measured flux for different near-Earth and space missions (Mrigakshi et al., 2012(Mrigakshi et al., , 2013. In particular, for solar maximum periods, the ISO15390 model resulted in considerable overestimation of all investigated particle fluxes for different missions, for example, over 70% for hydrogen nuclei in comparison with the measurements of different missions and that it underestimates the flux at low energy. It may thus be expected that this would lead to an overestimation of dosimetric quantities induced by GCRs, which is actually not the case in our study. For the case of solar minimum, the ISO15390 model has been reported to derive lower GCR hydrogen fluxes, in comparison with the data measured by different near-Earth and space missions and in comparison to more accurate models such as the Badhwar-O'Neill 2010 model (Mrigakshi et al., 2012(Mrigakshi et al., , 2013. Thus, for solar minimum, an additional underestimation of doses, due to intrinsic limitations of the input GCRs model, can be expected. Last, it is important to underline that neglecting heavier ions implies that the associated downward cascade and eventual backscattered radiation are not considered. The influence of heavy ions on different dosimetric quantities has been analyzed in several recent works. Röstel et al. (2020) found that, on the surface of Mars, the heavy ion primary particles contribute 9% of absorbed dose rate in a thin silicon slab phantom, while this ratio is slightly lower in a water sphere phantom, ∼8%. The heavy ion contribution to equivalent dose (not investigated here) was reported to be slightly larger, 11.5%. Other works Norman et al., 2016) investigated the influence of heavy-ion cosmic ray primaries at ∼30 km altitude at Earth, which has similarities with the Martian surface, and reported that the contribution of heavy ions to the absorbed dose and DE at that altitude is ∼20% and ∼50% respectively, though some overestimation of the quality factor possibly due to an over prediction of the flux was affecting the results . For solar maximum, considering only protons and α-particles should be of minor concern, likely because of the reduced mean free path of heavier ions in the atmosphere, the latter having a higher depth at the perihelion season than during early northern autumn in solar minimum 2009. However, with increasing atmospheric depth there would be an increase in fragmentation processes in the downward cascade, generating neutrons which give a relevant contribution in particular to the ADE, thus making the neglect of ions likely a critical aspect for the ADE, even in solar maximum. For solar minimum, it is clear that several different effects affect our results, and the neglect of high energy heavy ions is a considerable approximation (especially for the site at the lowest elevation, where the downward cascade would be more relevant), while RAD measurements are obviously sensitive to all particles. Nevertheless, here below we continue considering only protons and α-particles, similarly to previous studies Gronoff et al., 2015), and within such approximation we compare the results obtained at the different sites. As a last remark, as noted above, the definitions of the ADE and DE are different, and as also evident by the the discussion further in the text, also other works found ADE rates lower than the DE values for similar solar modulation conditions.  Table 2 only as an indicative reference. In principle, for October 2003, our surface absorbed dose appears to be slightly less than half of the absorbed doses recorded by MARIE, but both the planetary shielding effect, affecting the MARIE data as well as our results, and the fact that MARIE was covered by a 1.27 mm depth of aluminum and by a 100 μm depth of mylar thermal blanket (Zeitlin et al., 2010), thus not providing the fully unshielded dose in orbit, make it complicate to conclude on any impact of the sole atmospheric shielding effect on our results. The data from the TGO's cruise phase and the TGO orbits at high altitudes correspond to the declining phase of solar cycle 24. For high altitude orbits, the radiation conditions are very close to those in free space as shielding from Mars is negligible in such conditions. The difference between the cruise and orbit DE values can be interpreted mainly from the influence of the solar modulation (the cruise data are reported for April-September 2016, and the orbit data for November 2016-January 2017 and then February 2017-March 2017). Our ADE, for the declining phase of solar cycle 23, is ∼1/6 of the DE by Liulin-MO in the declining phase of solar cycle 24, which is reasonable as the measurements by the latter correspond to periods with a weaker heliospheric activity with respect to October 2003.

Radiation Spectra and Contribution from Different Particle Type
The overall radiation spectra (in terms of primaries, secondaries, and secondaries albedo, and in terms of downward and upward particles) do not show clear appreciable different trends among the three sites, and thus only the spectra for Oxia Planum are reported, while the contribution from different particle type is reported for both Oxia Planum and the location at Mawrth Vallis which differs more from Oxia in terms of regolith composition, that is Mawrth Vallis 2.
The spectra of the primary and (total and backscattered) secondary particles, summed over all particles types, at Oxia Planum, induced by GCRs protons and α-particles, are reported in Figure 3, for solar minimum and close to solar maximum during the declining phase of solar cycle 23 (early October 2003, before the rise of the October-November 2003 events). For both protons and α-particles, there is a relevant contribution of the backscattered, albedo component, caused by the interaction of primary particles and downward secondaries with the martian regolith, to the secondary flux. Also, the spectrum of the secondaries (both downward and albedo) generated close to solar maximum closely follows the one for solar minimum, with variable differences around 100 MeV/n and a reduced overall intensity for solar maximum at lower energies. The spectra of the secondaries are very similar for primary impacting protons and α-particles. This is reasonable as a substantial amount of 4 He ions fragment into protons in the atmosphere and follow the same energy loss processes as for protons (Guo, Saša, et al., 2019a).
In Figure 4, we report the downward and backscattered radiation spectra at the surface, separated per particle type, at Oxia Planum. Higher fluxes of the downward protons and α-particles, slightly shifted to lower energies, occur for solar minimum, as for solar maximum the stronger solar modulation leads to a reduction of the lower energy portion of GCRs. The high energy part of the spectrum is dominated by downward protons and neutrons, while neutrons and γ-rays are the predominant particles at the surface for energies lower than 100 MeV/n. The γ-contribution strongly decreases below the electron contribution below few tenths of keV. By comparing Figures 3 and 4, it can be easily noticed that, in case of impacting protons, protons reaching the surface are mostly primary particles, with the exception of those with energies below ∼10 MeV, which are likely generated in the interaction of the GCRs with the atmosphere.
In Figure 5, we plot the contribution of primaries, secondaries and particle type for Oxia Planum and Mawrth Vallis 2 to the ADE and ED hourly rate, for solar minimum and close to solar maximum, for impacting protons and α-particles separately. It can be observed that: (a) for the ADE, which is more sensitive to neutrons, primaries always contribute less than secondaries for both sites, and neutrons give the largest contribution for both sites and for both solar modulation periods; (b) for the ED, primaries contribute less than secondaries for both periods at Mawrth Vallis 2, suggesting the increased contribution of fragmentation secondary products to the ED; for Oxia Planum, primaries still contribute less than secondaries for solar maximum, but they contribute nearly to the same extent as secondaries for solar minimum; this suggests that at such higher elevation site more primary protons reach directly the surface, while for solar maximum the increased atmospheric depth likely still causes a considerable contribution from fragments. The ED is more sensitive to protons for proton impact and to neutrons for impact of α-particles, which is ascribed to the fact that α-particles have a smaller mean free path and are much more attenuated with depth; (c) there DA PIEVE ET AL.

10.1029/2020JE006488
10 of 20  is a change in the contribution of γ-radiation, linked to the efficiency with which different amount water in the regolith moderate neutrons. Neutrons always contribute more at Mawrth Vallis 2 compared to Oxia Planum, because of the increase in downward fragmentation processes due to the increasing atmospheric depth. However, as water is a good neutron moderator (Röstel et al., 2020), the more hydrated soil at Mawrth Vallis 2 reduces the number of albedo neutrons, especially below about 10 MeV. This influences the amount and type of further secondaries particles generated by the interaction of such moderated neutrons with the other mineralogical components of the regolith. Most γ-rays originate from such reactions, as neutron inelastic scattering and thermal or epithermal neutron capture. In the upward flux, the strongest γ-rays lines are produced by non-elastic-scattering processes such as 56 Fe (n,nγ) 56 Fe (0.85 MeV), 16 O (n,nγ) 16 O (6.1 MeV), 28 Si(n,nγ) 28 Si (strongest line at 1.8 MeV) and other higher energy γ-rays are produced by neutron capture processes such as 56 Fe (n,γ) 57 Fe (7.6 MeV) and 28 Si(n,γ) 29 Si (4.9 MeV). Looking at the γ-rays yield from the neutron capture 56 Fe (n,γ) 57 Fe at the two sites, being Fe present in slightly different amount at the two sites, we notice an enhancement of such yield at Mawrth Vallis 2. This is indeed interpreted as due to the fact that more moderated neutrons are available for such neutron capture reaction, and is in line with earlier studies (Masarik & Reedy, 1996). The overall contribution of backscattered γ-rays to the ADE hourly rate at Oxia Planum changes from 20% to 22.4% between solar minimum and maximum, while it stays constant at 25% at Mawrth Vallis 2. The variation of γ-rays at sites of astrobiological interest is important as γ-rays remain the reference radiation for the estimation of degradation of biological molecules (Dartnell et al., 2007;Ertem et al., 2017;Pavlov et al., 2012Pavlov et al., , 2002. Also, this may be relevant in future considerations of valuable sites for exploration-related resources on Mars with a mineralogy dominated by hydrated minerals. Here, the difference in the backscattered γ-rays remains nevertheless small, as the smaller amount of downward neutrons at Oxia Planum, combined with a higher Fe-content, also helps in lowering the contribution of neutrons to ground doses  and in giving rise to a similar γ-rays contribution. However, for the two sites considered here, the variation in Fe-content is only ∼4%, and in water-content is ∼10%, making water at Mawrth Vallis 2 a more efficient channel for neutrons modulation and following capture reactions by the Fe-oxide.

Dependence of GCR-Induced Doses on Diurnal Variations of Atmospheric Column Mass
Figure 6 (top panel) reports the day and night ADE and absorbed dose daily rates for Oxia Planum and Mawrth Vallis site 2, induced by GCRs protons and α-particles, for solar minimum and close to solar maximum, as a function of surface pressure. First, it can be noted that the pressure at each site is higher during the perihelion season (northern winter, October 2003) compared to the (early) autumn (January 2009), in agreement with the reports for different landing sites as a function of solar longitude (Martínez et al., 2017). The variation in the surface pressure however remains small at both sites, which is due to a non-optimal choice of the timing during day and night for which the calculations have been performed (2 a.m. and 2 p.m., not the peak-to-peak variation in the surface pressure daily cycle). Nevertheless, we observe that the day-night pressure variation for solar maximum is slightly higher compared to solar minimum, at least for Oxia Planum, as the solar forcing of tides increases when the amount of dust in the air rises (Leovy & Zurek, 1979;Medvedev et al., 2011). At Oxia Planum, the atmospheric column depth, directly linked to surface pressure from hydrostatic physics , varies from 20.37 g/cm 2 to 21.06 g/cm 2 for solar maximum, and from 18.08 to 18.56 g/cm 2 for solar minimum. Previous works have reported that, during northern winter, the increased dustiness of the atmosphere and the amplitudes of wave modes other than the sun-synchronous diurnal tide could tend to disrupt the diurnal tides propagation by modifying the atmospheric structure, reducing the pressure variation especially at low elevation sites with respect to higher elevation sites (Lee et al., 2009), which suggests that a smaller variation for Mawrth Vallis 2 could be expected.
A slightly higher surface pressure variation at Oxia Planum can also be expected looking at the vertical density profiles. These, as extracted from MCD ( Figure 6, bottom panel), suggest that at Oxia Planum, for solar maximum, the density between 120 and 160 km (upper inset) and at the surface (bottom inset) differs more between day and night compared to what occurs at Mawrth Vallis 2. The visible optical depth, which reflects the variation in opacity due to dust loading in this case, also suggests that at Oxia Planum a higher dust opacity is present compared to Mawrth Vallis 2, with a higher variation between the chosen day and night time (τ = 1.54 for vs. 1.60 for day and night respectively at Oxia Planum, and τ = 1.32 and 1.34 for day and night respectively at Mawrth Vallis 2). Overall, looking at the sole extent of the variation of doses and of atmospheric depths, we observe that the (day/night averaged) atmospheric depth differs more between the two sites for solar maximum compared to solar minimum (by 3 g/cm 2 for solar maximum and 1.8 g/cm 2 for solar minimum); however, the ADE and the absorbed dose differ more, among the two sites, for solar minimum. This suggests that even a small difference in the averaged atmospheric column depth among the DA PIEVE ET AL.

10.1029/2020JE006488
12 of 20 two sites plays a relevant role in solar minimum, contributing in determining the efficiency of the shielding towards the low energy portion of the GCRs, unmodulated during solar minimum.
The results in Figure 6 (top panel) suggest an anti-correlation of the ADE and absorbed dose with surface pressure for both sites for solar minimum, in agreement with a previous analysis based on RAD data (Guo et al., 2017). Such anti-correlation can be well expected as, for lower surface pressure, despite the fewer neutrons generated in the downward cascade with such thinner atmosphere, more primary impacting ions can survive traversal, resulting in a measured higher dose rate. As the total column depth increases, the Martian atmosphere, despite being very thin, acts as a shielding layer against the incoming GCR flux, in particular the low energy portion of the flux which is more easily shielded by the atmosphere and which is more abundant in solar minimum. For the absorbed dose, we find a variation of 8 μGy/day out of a daily absorbed rate of ∼230 μGy/day for Oxia Planum, and of 6 μGy/day out of a daily absorbed rate of ∼193 μGy/ day for Mawrth Vallis 2, in anti-correlation with surface pressure, similarly to the ADE. For the period close to solar maximum, we still find anti-correlation between the ADE day and night daily rate with surface pressure, which is in agreement with previous modeled results (Guo et al., 2017). The higher difference in the ADE day/night daily rate at Oxia Planum for solar maximum is due to the slightly higher difference in the density vertical profile compared to Mawrth Vallis 2. For the absorbed dose, we find an extremely weak anti-correlation of the absorbed dose rates with surface pressure for solar maximum, with a variation of only 2 μGy/day out of a daily absorbed rate of ∼110 μGy/day at Oxia Planum and a similar variation and absorbed dose for Mawrth Vallis 2. This is not in agreement with the correlation between absorbed dose and surface pressure expected on the basis of previous extrapolations for high solar modulation potentials (Guo et al., 2017), but the nearly coinciding absorbed dose rates suggest nevertheless a change in the dependence upon surface pressure.
Our lower estimation of the variations of the ADE and absorbed dose rates upon surface pressure with respect to RAD measurement is possibly also due to an amplification of the daily tide at Gale crater due to its peculiar topography. The variation of the absorbed dose rate measured by RAD over a diurnal cycle, which is ∼15 μGy/day peak to peak out of around 220 μGy/day for the period 2012-2014 (max Φ for this period 750 MV, with April 2014 maximum of solar cycle 24 [Guo et al., 2015]), corresponds to a daily pressure variation of ∼50 Pa as detected by the Rover Environmental Monitoring Station (REMS). Such variation is significantly larger than what observed at other landing sites on Mars and then what simulated for the Curiosity site by Global Circulation Models (GCMs) based on MCD -an averaged climate database. This is likely due to hydrostatic adjustment flows that amplify the daily tides and that are induced by the peculiar topography of the crater (Haberle et al., 2014;Richardson & Newman, 2018). However, when using actual GCMs with variability (Neary & Daerden, 2018) or mesoscale numerical models with sufficiently high horizontal resolution (grid spacing less than 5 km), then the large REMS daily pressure range are reproduced. Last, it should also be noted that the MCD atmospheric data used in this work inherently also have the seasonal CO 2 cycle, making impossible to distinguish the contribution to dose variation from the changes in atmospheric mass due to the CO 2 condensation cycle.

Scenarios for Short and Long Human Stays in Arabia Terra
The results in Table 3 consider two mission scenarios, one with a long stay on the surface of the planet (1year) during solar minimum and close to solar maximum, and the other one being a short stay (30-days), during which we consider the October 28, 2003 event. We only report the results for Oxia Planum, as the results for the other sites do not change the conclusions with respect to the limits established for stochastic and deterministic effects. The estimations for a long stay at Oxia Planum provide support for the future measurements by the Liulin-ML detector on the ExoMars 2022 surface platform.

GCRs Contribution During a 1-Year Stay-Solar Minimum and Close to Solar Maximum
For a 1-year stay, scaling up to an entire year the contribution from GCRs, our results suggest that the ED for solar minimum, 368.4 mSv (see Table 3), is nearly the double of the ED close to solar maximum, 190.8 mSv.
The ED limits for stochastic effects established by NASA (National Council on Radiation Protection and Measurements, Recommendations of Dose Limits for LEO, 2000) (620-1470 mSv for male and 470-1120 mSv for a female, for increasing age at initial exposure) are not overpassed, for none of the two solar modulation periods. However, the ED is a considerable portion of the limit for young women. Considering that most of the exposure is during the trip to and back from Mars, additional shielding is obviously necessary to maintain dose levels at the Martian surface as low as possible (Röstel et al., 2020).
The obtained ADE in Table 3 (131.8 mSv/year and 230.4 mSv/year, respectively close to solar maximum and close to solar minimum) are close to the DE from the updated extrapolation from RAD data for the exact solar maximum and solar minimum, with a worsening of the agreement for solar minimum. The result for solar minimum is higher than a previous estimate by dMEREM for Mawrth Vallis (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012), as it can be reasonably expected given the period considered in that work (2006, not exactly solar minimum) and the lower elevation of the specific location considered, but smaller than previous estimations in Simonsen, 1997(L. Simonsen, 1997. However, a good agreement is obtained with the DE obtained for solar maximum from the same author (L. Simonsen, 1997). Several previous studies (Cucinotta et al., 2001;Ehresmann et al., 2011) reported higher values during solar minimum, in particular an annual GCRs-induced DE ranging from 200-300 mSv/year, depending on the altitude.
The yearly absorbed dose for a water slab of 10 cm at Oxia Planum is 84.0 and 38.7 mGy/year respectively for solar minimum and close to solar maximum (scaled up from the corresponding daily rates). As said above for the average daily rates, both these values are considerably smaller than the values obtained from DA PIEVE ET AL. Note. Values for GCRs are obtained by scaling up the average daily values. Comparison is with the updated extrapolation from 7+ years of RAD data, other results as computed using HZETRN (L. C. Simonsen & Nealy, 1993;L. Simonsen, 1997), PLANETOCOSMICS (Gronoff et al., 2015) and dMEREM (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012). The results from (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012) for a 30-day stay only in presence of GCRs have been extracted from Table 6 in (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012) considering only protons and α-particles and for a period of 30 days assuming a constant distribution of GCRs flux. Uncertainty associated to dMEREM results is ∼1%. Abbreviations: ADE, Ambient Dose Equivalent; GCR, Galactic Cosmic Rays; HZETRN, High charge (Z) and Energy TRaNsport; SEP, Solar Energetic Particles.

Table 3 Absorbed Dose (mGy), ADE/DE (mSv), and ED as Induced by GCRs (Protons and α) for a 1-Year, a 30-Day Stay During Quiet Times, and Short Stay Scenarios
With an Occuring SEP Event the updated extrapolation from 7+ years of RAD data, which is likely due to a mix of factors mentioned above (modulation conditions, choice of the GCR model, and neglect of the heavy ions). On the basis of previous studies Röstel et al., 2020), however, the sole neglect of the heavier ions (and consequent generated downward cascade and backscattered radiation) would likely have a smaller impact on the absorbed dose results compared to the ADE.

GCRs and SEP Contribution During a 30-Day Stay Close to Solar Maximum
For a short 30-days stay during a period close to solar maximum, we find a GCRs-induced ADE which is higher than a previous estimate of the DE with the High charge (Z) and Energy TRaNsport (HZETRN) code (L. C. Simonsen & Nealy, 1993) for the solar maximum in 2014. This can be reasonably expected as in (L. C. Simonsen & Nealy, 1993) only atmospheric attenuation was taken into account but not the interaction of particles with the martian regolith, which leads to backscattering of neutrons and other secondary particles. Our ADE results are also higher than the results by dMEREM for solar maximum 2002 at Mawrth Vallis (a slightly different location in this region was considered) derived for a 30-day period from (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012), likely because the period of October 2003, although close to maximum, is in the declining phase of the solar cycle 23.
The obtained peak absorbed dose rate for SEP protons from the October 28, 2003 event is 3.43 mGy/day. The influence of dust loading on the value of the peak absorbed rate can be estimated by comparing with the value obtained by using the baseline MY26 scenario, which has a visible optical depth for Oxia Planum τ = 0.78 (τ = 0.67 for Mawrth Vallis 2). The change in the absorbed dose at the ground is rather small, 6% at Oxia Planum. This variation is similar, although smaller, to the results from previous studies on different SEP events in different applicable dust loading conditions for sites in the southern hemisphere (Norman et al., 2014). In order to assess the risk for deterministic effects, the SEP protons-induced accumulated absorbed dose in a 10 g/cm 2 water target proxy for BFOs (obtained by multiplying the hourly averaged dose rate by 1 h and then summing all the doses for the full duration of the event [Chenglong et al., 2008], with a total of 5.06 mGy) has to be scaled by the Relative Biological Effectiveness (RBE) of protons (1.5 [Operational Radiation Safety Program for Astronauts in Low-Earth Orbit: A Basic Framework. Bethesda, Md., 2002]).
This yields exposures far below even the smallest of the 30-days exposure limits, which is 250 mGy-Eq for the heart and BFOs (Guidance on Radiation Received in Space Activities. Bethesda, Md., 1989; National Council on Radiation Protection and Measurements, Recommendations of Dose Limits for LEO, 2000). We can therefore infer that for the (sole) October 28, 2003 event the health risks to BFOs and the heart were minimal if explorers experienced these events with no/minimal shielding, despite its higher fluence compared to the relevant September 2017 event detected by RAD since landing  (about one order of magnitude higher [Cohen & Mewaldt, 2018]).
Nevertheless, the series of events in October and November 2003 was very complex, with the onset of repetitive shocks lasting from October 19 through November 4, 2003 , with 10 shocks observed at the Advanced Composition Explorer (ACE) spacecraft and 5 CMEs (Lario et al., 2005  considered for late stochastic effects, to be added to the contribution induced by background GCRs. It is interesting to consider how the contributions of SEP primaries, secondaries and of particles type to the ADE and ED hourly rate during the SEP event, reported in Figure 7, change with respect to the GCRs-induced contributions in Figure 5. First, we notice that the ADE hourly rates are ∼22-24 times the ADE hourly rate induced by GCRs protons. The ADE hourly rates induced by the SEP peak flux differs more among the two sites compared to what occurs for background GCRs-induced ADE hourly rates, as the atmosphere modulates in a stronger way the lower energy SEPs flux, while the different amount of atmosphere to be traversed for the different sites is of lower importance for the transport of GCRs, especially for the (unmodulated) high energy portion. Close to solar maximum, primaries contribute less than secondaries to the ADE hourly rate, as it happens for background GCRs in early October 2003 ( Figure 5), while for the ED, primaries contribute more than secondaries, for both sites, contrarily to the case of GCRs background radiation. This change too is interpreted as due to the fact that the low energy SEP proton flux is more strongly modulated by the atmosphere with respect to the background high energy portion of GCRs (unmodulated during solar maximum), and the protons generated in the fragmentation of such low energy SEP flux portion contribute to the ED more than secondaries. Last, similarly to the case of GCRs, we observe that a slighly higher contribution of overall neutrons is present at Mawrth Vallis 2, again induced by a more relevant generation of the downward cascade, and that also the relative contribution of γ-rays remains slightly higher at such site.
Some previous works have reported the peak dose rate and the accumulated dose for the (different, or multiple) events of the Halloween storm at different locations (Jiggens et al., 2014;Joyce et al., 2015;Kozarev et al., 2010;PourArsalan et al., 2010;Schwadron et al, 2010). Importantly, near Earth, for the spectra of a 1-in-20-year event from SEPEM (Jiggens et al., 2012) closely agreeing with that of the October 28, 2003 event, the accumulated dose was reported to be ∼0.2 Gy-Eq (133mGy) (Jiggens et al., 2014) behind 1 cm (linear thickness) Al-shielding, close to the BFO limit 0.25 Gy-Eq. As the weakest region of an Extra-Vehicular Activity (EVA) suit is 15 times lower than this Al equivalent, effective warning for astronauts on EVA at Earth would have been fundamental (Jiggens et al., 2014). Since the interplanetary space dose rates could be considered to represent the possible upper limit of the radiation exposure in which an astronaut is doing EVAs when a SEP event occurs , we can therefore easily conclude that the BFOs limit would have been overpassed at Mars for the October 28, 2003 event, similarly to conclusions from previous studies on the October 26, 2003 event . While the atmosphere stops most of the SEPs, especially the low-energy ones, the latter do contribute greatly to the unshielded deep space dose.

Conclusions
We have presented a Monte Carlo particle transport study of the radiation environment and doses at Oxia Planum, the landing site for ExoMars 2022, and at two locations at Mawrth Vallis, previously under scrutiny as candidate landing sites respectively for the ExoMars 2022 and the current MSL mission, during solar minimum and close to solar maximum during the declining phase of solar cycle 23, using the ESA's official model dMEREM. The dMEREM standalone version available to users, via ESA's permission, was used and some post-processing tools were implemented to allow to use of different vertical profiles from MCD and to calculate the absorbed dose (not automatically printed by dMEREM). The results may serve as reference for future measurements from the Liulin-ML detector on the ExoMars 2022 surface platform and point to some differences with respect to the results from previous analysis of RAD data for Gale crater.
The radiation environments among the three sites appear, overall, similar. The comparison of the daily ADE and absorbed dose rates with the updated extrapolation from RAD data highlights the importance of some differences in the input modulation conditions, especially for solar minimum, considered in the present calculations and in the analysis of RAD data, intrinsic limitations of the ISO15390 model in describing accurately the flux of primary light ions, and the neglect of heavier ions in our work which may affect both the results for solar minimum and those for solar maximum. Small differences are found in the γ-rays contribution at Oxia Planum and Mawrth Vallis site 2. In particular, neutrons contribute more to the ADE rates at the deepest Mawrth Vallis site 2, due to increased fragmentation in the downward cascade. However, in the backscattered components, a ∼10% increase of water content at Mawrth Vallis 2 with respect to Oxia Planum moderates more efficiently fast neutrons. With more moderated neutrons, γ-rays in the backscattered component generated in neutron capture reactions are slightly enhanced. The higher water content at Mawrth Vallis and the consequent increased neutron modulation overweighs the eventual decreasing of neutrons solely given by Fe-oxides, given the small difference in hematite content at the two sites. The fact that γ-radiation may increase in presence of a more hydrated soil should be considered when considering the capability of minerals/materials to shield from radiation at specific sites, in the study of the shielding of biological molecules from radiation by mineral matrices, and in the consideration of possible scenarios for ISRU activities at sites with a mineralogy dominated by hydrated minerals.
In studying the dependence of doses on surface pressure, we only obtain a partial agreement with previous extrapolations from RAD data for different solar modulation conditions. An anti-correlation between ADE and absorbed dose day/night daily rates and surface pressure is found for both solar minimum and maximum. For solar minimum, this is in line with . For solar maximum, we only observe a weakening of such anti-correlation, or actually nearly coinciding dose rates, for both sites, contrary to the correlation reported by previous works. A slightly higher day/night pressure variation and ADE day/night daily rate variation at Oxia Planum could be related to a higher difference in the density vertical profiles above Oxia Planum in the dusty northern winter close to solar maximum. The present results suggest that not all the trends in the variation of doses reported by RAD could be similarly reproduced for the nothern hemisphere in the different seasons, and for sites with different dust loading.
The comparison of the results for a long (1-year) and short (30-days) surface stay suggests reasonable agreement with previous results from dMEREM (McKenna-Lawlor, Gonçalves, Keating, Morgado, et al., 2012), and for solar maximum with HZETRN (L. Simonsen, 1997) and PLANETOCOSMICS (Gronoff et al., 2015). The atmosphere of Mars provides a sufficient amount of shielding from the October 28, 2003 event, and incurred doses are below permissible limits, despite the considerably higher fluence with respect to the relevant September 2017 event detected by RAD. However, in reality the Halloween event featured multiple interplanetary shocks and ICMEs , which should be taken into account for late stochastic effects, in addition to the contribution from background GCRs. The influence of dust loading induces a variation with respect to the baseline MY26 scenario similar to previous studies (Norman et al., 2014).
Further work is needed to better discern the causes of the differences with the data from RAD at Gale crater, in particular by adding heavier ions and also possibly disentangling seasonal effects, that is effects of the CO 2 condensation cycle. At the same time, additional work would be needed to investigate the impact of a higher simultaneous variation of water and Fe-content for sites at different elevation on the doses at the surface and, possibly, in the subsurface.

Data Availability Statement
The data relative to Figures 2-7 can be found at the Zenodo platform (Da Pieve et al., 2020, figure data for "Radiation environment and doses on Mars at Oxia Planum and Mawrth Vallis: support for exploration at sites with high biosignature preservation potential" Zenodo, 10.5281/zenodo.3779187). The GCR spectra in Figure 1 can be retrieved from the SPENVIS interface (https://www.spenvis.oma.be/) and the differential profile fluxes for the October-November 2003 period from the SEPEM tool (http://sepem.eu/). and Technology Centre, ESEA, the Netherlands) for providing a standalone version of dMEREM, Dr. Ian Thomas (Royal Belgian Institute of Space Aeronomy, Belgium) for help with the SPICE-enhanced version of the visualization tool Cosmographia (https:// naif.jpl.nasa.gov/naif/cosmographia. html) and Dr. Vincent Letocart (Royal Belgian Institute for Space Aeronomy, Belgium) for assistance for computational aspects. J. G. is supported by the Strategic Priority Program of the Chinese Academy of Sciences (Grant No. XDB41000000 and XDA15017300), the National Natural Science Foundation of China (Grant No. 42074222) and the CNSA pre-research Project on Civil Aerospace Technologies (Grant No. D020104). The work has received fundings from the Research Executive Agency under the EU's Horizon 2020 Research and Innovation program (grant ID 776410). This work benefited from networking activities carried out within the EU funded COST Action CA17126 (TUMIEE) and represents a contribution to it. G. Gronoff and C. J. Mertens were supported by the Advanced Exploration Systems Division within the NASA Human Exploration and Operations Mission Directorate. We acknowledge the RAD data which are archived in the NASA planetary data systems' planetary plasma interactions node (http://ppi.pds.nasa.gov/).