atmosphere SPICAM on Mars Express: A 10 year in-depth survey of the Martian Icarus

ContentslistsavailableatScienceDirect

Icarus

journalhomepage:www.elsevier.com/locate/icarus

SPICAM on Mars Express: A 10 year in-depth survey of the Martian atmosphere

F. Montmessin

^a,∗

, O. Korablev

^b

, F. Lefèvre

^c

, J.-L. Bertaux

^a

, A. Fedorova

^b

, A. Trokhimovskiy

^b

, J.Y. Chaufray

^a

, G. Lacombe

^a

, A. Reberac

^a

, L. Maltagliati

^d

, Y. Willame

^e

, S. Guslyakova

^b

, J.-C. Gérard

^f

, A. Stiepen

^e

, D. Fussen

^f

, N. Mateshvili

^f

, A. Määttänen

^a

, F. Forget

^g

, O. Witasse

^h

, F. Leblanc

^a

, A.C. Vandaele

^f

, E. Marcq

^a

, B. Sandel

ⁱ

, B. Gondet

^j

, N. Schneider

^k

, M. Chaffin

^k

, N. Chapron

^a

aCNRS LATMOS, Laboratoire Atmosphères, Milieux, Observations Spatiales, Université Versailles St Quentin en Yvelines, Quartier des Garennes, 11 bd d’Alembert, 78280 Guyancourt, France

bSpace Research Institute, 84/32 Profsoyuznaya St, Moscow, Russia

cCNRS LATMOS, Université Pierre et Marie Curie, 4 place Jussieu, 75252 Paris, France

dCEA-Saclay, route Nationale, Gif-sur-Yvette, France

eBIRA-IASB, avenue Circulaire 3, Brussels, Belgium

fLPAP, Université de Liège, Quartier Agora, 19c allée du 6 août, Liège, Belgium

gCNRS LMD, 4 place Jussieu, 75252 Paris Cédex, France

hESA-ESTEC, Keplerlaan 1, 2201 AZ Noordwijk, Netherland

iLPL, University of Arizona, Tucson, AZ, United States

jInstitut d’Astrophysique Spatiale, Orsay, France

kLASP, Boulder, CO, United States

a r t i c l e i n f o

Article history:

Received 14 December 2016 Revised 11 May 2017 Accepted 14 June 2017 Available online 22 June 2017

a b s t r a c t

TheSPICAMexperimentonboardMarsExpresshasaccumulatedduringthelastdecadeawealthofob- servationsthat haspermitteda detailedcharacterization ofthe atmosphericcompositionand activity fromthenear-surfaceuptoabovetheexosphere.TheSPICAMclimatologyisoneofthelongestassem- bledtodatebyaninstrumentinorbitaroundMars,offeringtheopportunitytostudythefateofmajor volatilespeciesinthe Martianatmosphereoveramulti-(Mars)yeartimeframe. Withhisdualultravio- let(UV)-nearInfraredchannels,SPICAMobservesspectralrangesencompassingsignaturescreatedbya varietyatmosphericgases,frommajor(CO₂)totracespecies(H₂O,O₃).Here,wepresentasynthesisof theobservationscollectedforwatervapor,ozone,cloudsanddust,carbondioxide,exospherichydrogen andairglows.TheassembledclimatologycoverstheMY27–MY31period.However,themonitoringof UV-derivedspecieswasinterruptedattheendof2014(MY30)duetofailureoftheUVchannel.ASO₂de- tectionattemptwasundertaken,butprovedunsuccessfulfromregionaltoglobalscales(withupperlimit greaterthanalreadypublishedones).Oneparticularconclusionthatstandsoutfromthisoverviewwork concernstheway theMartianatmosphereorganizesanefficient masstransferbetweenthelower and theupperatmosphericreservoirs.Thishighwaytospace,aswenameit,isbestillustratedbywaterand hydrogen,bothspecieshavingbeenmonitoredbySPICAMintheirrespectiveatmosphericreservoir.Cou- plingbetweenthetwoappeartooccuronseasonaltimescales,muchshorterthantheoreticalpredictions.

© 2017TheAuthors.PublishedbyElsevierInc.

ThisisanopenaccessarticleundertheCCBY-NC-NDlicense.

(http://creativecommons.org/licenses/by-nc-nd/4.0/)

1. Introduction

The SpectroscopyforInvestigation ofCharacteristicsoftheAt- mosphere of Mars (SPICAM) instrument is a dual ultraviolet in-

∗ Corresponding author.

E-mail address: franck.montmessin@latmos.ipsl.fr (F. Montmessin).

fraredspectrometerthatwasdesignedto retrievetheabundances ofasetofmajorandminorspeciesoftheMartianatmosphere.On- boardtheMarsExpressmission,SPICAMhashadthecapabilityto performobservationswithavarietyofgeometricalconfigurations;

monitoringthecolumn-integratedabundancesofozone,waterva- porand aerosolsin a nadir-looking mode, aswell astheir verti- cal distribution between ∼10 and 150km using stellar and solar http://dx.doi.org/10.1016/j.icarus.2017.06.022

0019-1035/© 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license. ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

Fig. 1. A compilation of five Martian years of observations by SPICAM in a nadir looking mode. The figure displays zonally averaged values as a function of solar longitude (Ls) of the retrieved (from top to bottom) water vapor, ozone, molecular oxygen singlet delta daytime emission in the near infrared, and dust as well as water ice opacity at 250 nm. The degradation of the UV channel has prevented reliable derivation of ozone and dust/clouds in the ultraviolet during MY31. The unit for each quantity, whose scale is plotted on the right hand-side, is given in brackets.

Note the high degree of repeatability in the seasonal behavior of all quantities, to the exception of dust during southern summer. Note also that displayed abundances have been averaged out of a variety of local times, from morning to afternoon.

occultations. In a dedicated limb staring mode where the Mars Expressspacecraft remains ina fixed inertial position,SPICAM is used to infer the density ofhydrogen atoms from 200 up to 10 000kmofaltitudewhilerelyingontheresonantlyscatteredsolar photonsattheLyman-alphaemissionlineat121.6nmofHydrogen atoms.Thissamemodeofobservationisusedfortheidentification ofa varietyof airglows, some ofthem beingassociatedwith the presenceofauroras,asdetectedforthefirsttimebySPICAMabove themagneticfieldcrustal anomalies(Bertauxetal., 2005a).Since MarsExpress orbitinsertion on December 23,2003, SPICAM has operated continuously, accomplishing one of the longest surveys todate ofthe Martianatmosphere(see Fig. 1),pursuing the ref- erenceclimatologyestablishedbytheMarsGlobalSurveyor(MGS) missionthatstartedneartheendofthe20thcenturyandpaving the way to the Imaging Ultraviolet Spectrograph (IUVS) onboard theMarsAtmosphereandVolatileandEvolution(MAVEN)thathas beenoperatingsinceSeptember2014.

Thepurposeofthispaperistoprovideanoverviewofthemain scientific results that have been collected thanks to the multi- annualmonitoring madeby SPICAM.Several resultsdescribed in thismanuscriptare presented ingreater depth in companionar- ticles that specifically address a particular theme of the SPICAM observations(Trokhimovskiyetal.,2015;Lefèvreetal.,inprepara- tion,2017;Guslyakovaetal.,2016).Thisarticleisorganizedaround thefollowingtopics:

• Lower atmosphere sounding including observations and map- pingofwatervapor,ozone,O₂ andaerosols.

• MiddleatmospheresoundingwithacharacterizationoftheCO₂, O₂,O₃andaerosolsfrom50to150kmaswellasNOemissions subsequenttoNandOatomsrecombination.

• Upperatmospheresoundingtoinfertheconcentrationofhydro- genandoxygenatomsabovetheexobase.

2. Instrumentstatus 2.1. Instrumentdescription

The SPICAM set-up comprises separate ultraviolet and near- infraredchannelsin oneoptical unit bothconnectedto thesame Data Processing Unit. For the interested reader, detailed techni- cal descriptions of theinstrument canbe found in Bertauxetal.

(2006)andKorablevetal.(2006a).

The ultraviolet channel uses an optical entrance of 4cm di- ameterfeedingan off-axisparabolicmirrorfocusing theobserved scene at the focal plane. The instantaneous field of view (IFOV) is limited atthe focal point ofthe parabolic mirror by a 50

μ

^m

wideretractableslitthatextendsoveranangularapertureof2.8°, equivalentto about two pixels on the UV detector. Inthe upper mostportionoftheslit,a10timeswideraperture allowsforthe observations of fainter sources at the expense ofdegraded spec- tral resolution. Spectral dispersion is accomplished by a concave holographictoroidalgratingimprintingthediffractedordersatthe entrance of an MgF2 window covering a solar blind CsTe pho- tocathode with zero quantum efficiency beyond 320nm. An ad- ditional sapphirefilter is glued above the window and covers it

150 200 250 300 Wavelength (nm) 0.0

0.2 0.4 0.6 0.8 1.0

Voigt Coefficients

150 200 250 300

Wavelength (nm) 1.0

1.2 1.4 1.6 1.8 2.0 2.2 2.4

Voigt FWHM (nm)

Fig. 2. (Left) Wavelength-dependent parameters of the Voigt profile H(a,u) that best fits the SPICAM UV channel Point Spread Function. The solid line refers to the damping parameter a while the dashed line is for the offset u/x. (right) Estimate of the full width at half maximum of the Voigt profile and its dependence to wavelength. The brutal variation above 250 nm is due to the effect of the sapphire window glued at the entrance of image intensifier, combined with the rapidly changing index of refraction of MgF2 at short wavelength. The absence of the window (pixels > 270) improves the focusing in the remaining range.

partially,preventingoverlappingofdiffractionordersandLyman-

α

straylight. Thisdefinestheentranceofthe imageintensifier that consistsofaMicroChannelPlate(MCP)whosegainisadjustedby a high voltagelevel ranging from500 to 900 V. The CCD detec- tor is a Thomson TH7863 with 288 × 384 pixels (23

μ

^{m pitch)}

cooled by a thermo-electric device ensuring lower dark current.

Thespectralresolutionoftheinstrumentiswavelengthdependent;

it exceeds1.25nm over theentirerange, increasingup to 2.3nm around260nm(seeFig.2).

ThefocallengthoftheUVtelescopeissuchthatoneCCDpixel covers a IFOVof0.01°× 0.01° Thenarrow partofthe slitofthe spectrometer is 0.02° wide by 1.9° long while the wide part is 0.2°wideby0.98°long.Toreduce thetelemetryburden,onlyfive binned areas of the detector are transmitted at the end of each acquisition. Thesebinnedareasconsist ofsummedlines ofpixels that correspond to contiguous areas in the observed scene. It is thus possible to obtain the spatialvariation ofthe UV spectrum alongtheaxisoftheslit.Thenumberoflinesperbinisdefinedat commandingandisusuallychosen toaccommodatetheexpected brightnessoftheobservedscene.Forapointsourcesuchasastar, the standard binning covers 16 lines of pixels while for an ex- tended sourcesuchastheMarssurface,thebinningisreducedto 4linesonthedetector.

The near-infraredchannel ofSPICAMhas inauguratedthefirst use of acousto-optic tunable filters (AOTF) in deep space. Built by IKI, the SPICAM near-IR channel accommodates a relatively high-resolution spectrometerinaverylightdesign(<800g).The AOTF operating principle relies on producing interferences be- tween the incident optical beam and an acoustic wave in a bi- refringent medium(TeO₂). Theinput optical beamsustainsa vol- umediffractionthatisolates twodiffractionordersaway fromthe un-diffractedfractionoftheinputbeam. Theacousticwaveprop- agation insidethe TeO₂ crystalis generatedby a radio-frequency (RF) controllerthat synthetizesthefrequencyatwhichthetrans- ducer mustbe excited to createan acousto-opticinterferencein- side the crystal. Because there is a strict relation between the input RF frequency and the diffracted wavelength, one is able to retain a specific wavelength selected within a range equiva- lent to an octave of the RF controller. Because the energy re- mains concentrated within the first (+ 1 and −1) diffracted or- ders,suchdevicesdonotrequireadditionalfilterstorejecthigher diffraction orders. This leads to a simple and adaptable spec- tral analyzer (any wavelength can be accessed instantaneously).

ThereisnoslitintheAOTFspectrometer,increasingthethrough- put compared to traditional grating spectrometers. In turn the

AOTF relies on a progressive scanning of the wavelengths of in- terest to reconstructa given spectrum thus reducing the overall efficiency.

Withthisset-up, thenear-IR channel ofSPICAM iscapable of isolatingateachacquisitiona0.5–1.2nmbandwidthwithintheac- cessible1to 1.7

μ

^{mrange.In therangeofthe watervaporband}

the resolving power

λ

^/

λ

^is ∼2000. Wavelengths are registered sequentiallywiththeincrementof∼0.3×

λ

^{windowingorloose}

samplingallows reducing the duration ofmeasurements. To bet- tercharacterizeatmosphericaerosols,thecontinuousabsorptionis alsomeasuredatseveraldistantwavelengths.TheIFOViscircular, andamountsto1°fornadirmeasurements,andto∼4whenob- servingtheSun.

2.2.Calibrationandinstrumentageing

As explained in Bertauxet al.(2006),we define the sensitiv- ityoftheSPICAMUVchannelasthesensitivityoftheinstrument fora stellar source placed atthe center ofthe spectrometer slit.

The nominalwavelengthassignment is definedaccordingly,mak- ing it validfor all extended source measurements collectedwith the slit in place. The so-called efficient area S_eff (in cm²) is a wavelength-dependentfunctionoftheactualinstrumentphotosen- sitive area collecting the photonsemitted by a point source(see Fig.3).ThedisplayedS_effcurveisobtainedwhenthestarsignalis integratedover 16CCD lines(centered onCDD line144) encom- passingroughly90%ofthetotalsignal(theremaining10%spread- ingawayfromthedispersionarea).Thedropatshortwavelength cut-off isdue to the opacity of MgF₂ window, while thedrop at longerwavelengthreflectsthedrop-off ofthephotocathode.

An updated version of the calibration procedure of the SPI- CAM UV spectra is given in Snow et al. (2013). We recall here themainexpression tobe usedforconvertingSPICAMUV detec- tor readouts(in Analog-to-Digital Units,ADU) into physical units (photons×cm⁻²)forapointsource:

N_ADU/pixel/acquisition=g×Np×

λ

^×Ti×S_{e f f}×fstar

andphotons×cm⁻²×sr⁻¹forextendedsources:

NADU/pixel/acquisition=g×Np×

λ

^×Ti×Se f f×

ω

The number of ADUs N_ADU yielded by a pixel during one in- tegration is therefore proportional to the gain of the detector g, thenumberNp ofphotonsdetected per wavelengthunit andper second,thespectral sampling

λ

⁽∼0.54nmper pixel)andtothe integrationtime T_i(insecond).Incaseofanextendedsource,the

Fig. 3. Updated version of the efficient area Seffof the SPICAM UV channel based on the updated calibration procedure established by A. Reberac and reported in Snow et al. (2013) . The previously published version of Seff(see Bertaux et al., 2006 ) is shown for comparison in dashed line.

solidangle

ω

^{subtendedby a pixel(8} × 10⁻⁸sr)needs tobe in- cluded as an additional multiplicative term. The parameter fstar

referstothefractionoftheincidentphotonsfallingexactlyonthe pixel.Asindicated by Bertauxetal.(2006), abinning of16 lines of pixel contains around 90% of the stellar light (fstar=0.9). The recordedevolutionoftheinstrumentsensitivityisshowninFig.4 forthe10yearsoftheSPICAMsurvey.Theplottedquantityisthe instrumentgain g asafunction of theintensifier highvoltage.It isclearthattheSPICAMUVchannelencounteredagradualdegra- dationfromlate 2011onwards thateventually led toa >10-fold reductionofsensitivity,stillallowingustomaintainacceptablesci- entificperformances.

2.3.Instrumentstatus

AsofDecember2014,SPICAMUVchannelhasceasedreturning sciencedata,sendingonlydarkframesinstead.Thissituationisthe outcomeofaprogressivedemiseoftheUV channel whoseorigin wasfirst identified 9 years before. In January 2006, the SPICAM UVchannel startedtoexhibit ananomalous image intensifierbe- havior,characterizedbymultiplesporadicandspuriouschangesof thehighvoltagesettingduringasequenceofobservation.TheSPI- CAMUV channeldatacleaning process haseventuallyallowed us toretainall healthyspectrawhileidentifyingmost(> 95%)ofthe affectedones.SPICAMLevel0Cdataintheultravioletavailableon theEuropean SpaceAgency’sPlanetaryScienceArchive havebeen producedthisway,eachrecordedspectrumbeingassociatedwith a flag parameter indicating whether the spectrum is infected or not.

For additionaldetails, the readerisinvitedto consult the Ap- pendixsectionwhereasummary isgivenofthemainchanges in dataprocessingandinourunderstanding oftheinstrumentchar- acteristicssince the firstpublication on thistopic(Bertauxetal., 2006).

TheIRchannel,ontheotherhand,hasexhibitedastablebehav- iorthroughouttheMEXmission,andnonoticeabledegradationof theIRspectraqualityhastobereportedatthispoint.

2.4.Dataset

Since thebeginning ofits operationsatMars, SPICAMhas ac- complishedmorethan12,500sequencesofobservations(seeTable 1) amongwhichseveralthousandsofatmosphericprofilesaswell asacontinuousmulti-annualtrackinginnadirmode.

3. Loweratmospheresounding(0–50km)

ContrarilytomostNASAmissions,theESAMarsExpressorbiter does not obey a sun-synchronous orbitand thus nadir mapping coversavarietyoflocaltimesfrommorningtoafternoon.Thereis a specific interestforidentifying andstudying thelocal time de- pendence of thespecies behavior (suchas O₃ andH₂O), butthe MarsExpressorbitdoesnotperformaregularlocaltimesampling andsomealiasingwithseasonaltimescaleisdifficulttoavoid.For thisreason,localtimedependenceisnotdiscussed inthefollow- ing yetshould be the purposeof futurein-depth analysisofthis dataset.

3.1. Watervapor

•Nadirmapping

The infraredchannel ofSPICAM hasprovideda nearly contin- uous survey of the column-integrated abundance of water vapor since2004usingthewatervapor1.38

μ

^{mabsorptionband(asde-}

scribed inFedorova etal., 2006a). Thisdataset spanningMartian Years27–31(seeFig.1)enablestheinvestigationofboththesea- sonalandtheinter-annualvariationsofatmosphericwater.Thera- diative transfer model used for the retrieval accounts for multi- ple scatteringcaused by dust and clouds.Comparedto the orig- inal version of the retrieval reported in Fedorova et al. (2006a), thedatatreatment(Trokhimovskiyetal.,2015)usedfortheresults presentedhereincorporatesanumberofimprovementsmainlyre- latedtoabettercharacterizationoftheinstrumentcalibrationand ofthemodelinputparameters.Thesensitivityofresultstoaerosol properties,surfacealbedo,solarspectrum,andwatervaporverti- caldistributionhasbeencarefullyevaluated.Forinstance,neglect- ing theaerosolscatteringbiasestheinferred wateramountbyas muchas60–70%duringdustyevents.Watervaporretrievalisalso sensitive tocloud activityespeciallyaround aphelionin thetrop- ics. However,theeffectofcloudsanddusthavebeenincludedin theretrieval.Informationondustandiceabundanceswereextrap- olated fromthepublishedTHEMIS dataset(whichspans untilthe middleofMY30-Smithetal., 2009)down tothenear-IRrangeof SPICAM.

SPICAM observations reveal a water vapor seasonal behavior that is to a large extent in agreement withprevious monitoring (inparticularTESdataset,asrevisedfromSmith,2004);watercol- umnintegratedabundancepeaksnear60–70pr.

μ

^minthenorth-

ernsummerpolarregionasthepolarlayereddepositsgetexposed to intensified spring/summer insolation whereas only 20pr.-

μ

^m

is observed during the southern summer at the south pole. The equivalentvolume oficesublimatingasvapor intheatmosphere isabout0.5km³ atthebeginning oftheyearandlater buildsup to2.3km³ atLs=100° atatime whenwatervapor abundanceis foundmaximumonaglobalaverage.

To the exception of the MY28 dust event that is discussed below, SPICAM reveals a water cycle that is stable on average throughout the terrestrial decade monitored. With the revisit of the Mars Atmospheric Water Detector dataset (Fedorova et al., 2010), the picture of a Mars’ water cycle evolving in a quasi- stabilizedstateappearsmoreandmoreconfirmed(Richardsonand Wilson,2002; Montmessinetal., 2004,2017). Mars’swatercycle respondsessentiallytothelargeseasonalandspatialvariationsof insolationoverthe globe.It ispossiblethat variabilitymayoccur ontimescalesgreaterthanadecade,butthecompilationofwater vapormeasurementsmadetodatesuggeststhewatercycledidnot changesignificantlyovermorethan20Marsyears.

When comparing Martian Years (Fig. 1), one is able to cap- ture the prominent dropof water abundance that occursduring the dust storm ofMY 28. This sudden decrease of water, which was alsoreported in CRISM data (Smith et al., 2009), cannot be

0 50 100 150 200 High voltage level 0

20 40 60

Gain

2004-2007 2008-2011 2011-2012 2013-2014

2004 2006 2008 2010 2012 2014 Years

-3 -2 -1 0 1 2 3 4

Ln(Gain)

Fig. 4. (Left) SPICAM UV channel intensifier gain curve as a function of the high voltage level (1–28 = 256). (right) Gain (in logarithmic scale) evolution with time for a fixed value of high voltage. The curves on the left distinguish between various periods during the mission, highlighting the relatively stable behavior from 2004 to 2011, and the significant loss of sensitivity that occurred in 2011, after which time, SPICAM intensifier lost a rough factor of 10 in efficiency. Near the end of 2014, SPICAM UV channel ceased to transmit observed signal, suggesting the loss of the intensifier.

Table 1

Breakdown of SPICAM observations against modes of operation as performed since the beginning of Mars Express operations until the end of 2014 when the UV channel stopped nominal operations.

Channel Nadir Limb Stellar occultation Solar occultation Phobos/deimos Calibration Sky Earth Comet Total

UV 3553 1724 4625 1096 238/67 866 261 5 19 12,454

IR 3549 1494 1855 1036 236/67 12 27 0 16 8292

attributedto measurementbiasescaused by shieldingof thewa- ter moleculesby dust. It appears to reflectan actual dropof at- mosphericwaternearorabovethesurfaceduringtheduststorm.

Dedicated radiativetransfercalculationsindicatethatwatervapor retrieval based on the optically thin 1.38

μ

^{m absorption band is}

mostlysensitivetothewaterpresentinthefirstkilometersofthe atmosphereabovethesurface.Wespeculatethattheobservedde- cline ofthewaterabundanceduringMY28is consistentwithen- hanced adsorption by the Martianregolith. The higher adsorbing capacity of the regolith underconditions of globallyspread dust storms has been predicted by Böttger et al. (2004, 2005) based on Global Circulation Model (GCM)computations andappears to becorroborated bythepresentmeasurements.Notethat cluesfor a diurnal“breathing” of theregolithata localscale couldnot be supportedbytheavailableSPICAMmeasurementsduetotheinad- equateMarsExpresslocaltimecoverage.

ItistobenotedthatwithinthecontextofMarsExpress,anim- portantendeavorofinstrumentcross-comparisonwasundertaken in order to understand the differences of retrieved water vapor columnabundances. TheObservatoire pourlaMinéralogie, l’Eau, les Glaces et l’ Activité (OMEGA) (Bibring et al., 2005; Encrenaz etal., 2005) andthePlanetary FourierSpectrometer(PFS) instru- ments (Formisanoetal., 2006; Fouchet etal., 2007) alsoprovide measurements of water vapor and these were compared to SPI- CAM (Tschimmeletal., 2008). Thiscomparisonstudyallowed for aquasi-convergenceofallMarsExpressinstrumentsyetsomede- partures (within20–30%) remainedthat theverynatureofobser- vations(e.g.spectralrange)couldpartlyexplain.Atthattime,SPI- CAMwasclearlystandingoutfromtheseothermeasurements,but an in-depthreanalysis led to realigning SPICAM with the restof MarsExpress(Trokhimovskiyetal.,2015).

•Verticalprofilesfromsolaroccultations

SPICAMdeliveredthefirstmonitoringofthewatervaporverti- cal profilebetween∼10and100kmduringa wholeMartianyear (Fedorovaetal.,2009;Maltagliatietal.,2011,2013).Thethreeoc- cultationcampaignspresentedhere, allowustoexamine asome- whatunexpectedbehaviorofwatervaporverticaldistribution(Fig.

5). Watervapor isfound lessabundantandmoreconfinedinthe

loweratmosphereduringnorthernspring/summer,anexpectedre- sultwhenconsideringthatwatervaporshouldremainsequestered below the condensation level (hygropause). However, as demon- stratedbypreviousSPICAMstudies,theverticaldistributionisnot entirelydictatedbythevaporsaturationconditions:largeexcesses ofsaturationhavebeenobserved,especiallyatmid-latitudesinthe northern hemisphere spring(Maltagliati etal., 2011). In contrast, thewarmersouthern springappearsassociatedwithahigherver- ticalmobilityforwatervapor(asillustratedbyFig.6),acharacter- isticthatwasobservedespeciallyafterLs =240°There,recurrent featuresofdiscretelayersofwatervaporwith>100ppmvmixing ratiohavebeenreportedabove60km.

Whilethewatervaporprofilesappeartoreflectthelargetem- poral and spatial variabilities of the Martian climate, they also show sensitivity to more localized and short-term meteorologi- cal events. Abrupt variations (within a day or so) of water va- porabundance andinverticaldistribution canbe identified. Wa- ter upsurges, appearing suddenly within few Martian days,were observed three times in MY29 as described in Maltagliati et al.

(2013) and reproduced here in Fig. 6. Similar short-term varia- tions affect the aerosol profiles, acquired by SPICAM simultane- ouslywithwater,suggestingthat transportphenomenaaffectwa- terandaerosolslikewise.Aerosolupsurgesareobservedsystemat- icallyinconjunctionofthoseoccurringwithwatervapor andde- tached layers ofaerosols are often coupled withthe waterones.

Theseobservationsdemonstratetheimportanceoftheverticaldis- tribution as crucial diagnostics for key processes in the Martian watercycleandclimateasawhole.

3.2.Ozone

•Nadirmapping

Between2004and2011,SPICAMobservationsinnadirviewing provided a continuous and near-complete coverage of the ozone columndistribution,one ofthelargestdatabaseofMartianozone to date. The ozone column is retrieved from the UV channel si- multaneouslywith themean(210–290nm) dustopacity andsur- facealbedo.Perrieretal.(2006)presentedaninitialversionofthe retrievalalgorithmtoprocessthefirst1.2MartianyearofSPICAM

Fig. 5. (Top left) Geographic and seasonal distribution of the three Mars years of infrared solar occultations compiled in this work. (Top right) maximum value of the water vapor volume mixing ratio detected for each orbit. Only the orbits with positive detection of water vapor are plotted. The orbits with a black circle exhibit a detached layer.

The arrows indicate the moments of the sudden upsurges of water (see text). (Bottom left) maximum value of the water saturation ratio for the orbits where supersaturation is detected. (Bottom right) hazetop, defined as the altitude where the aerosols’ optical depth drops to 1 from below. The black cirles identify orbits with detached layers, and the arrows mark the position of the upsurges.

Fig. 6. Compilation of all the derived values of the hygropause altitude, defined as the level above which water vapor can no longer be detected.

operation.Theirmethodusedafullradiativetransfermodelofthe Marsatmosphereandaccountedformultiplescatteringcausedby dustandclouds.Recently, theUV nadir datasethasbeenentirely

reprocessedwithanimprovedversionofthealgorithm.Thanksto abettercharacterizationoftheefficientareaS_effandtheimproved modeling of straylight in nadir spectra,the ozone retrieval now reliesontheabsoluteUVreflectance observedbytheinstrument.

Hence,itnolongerrequirestheuseofaso-called“referencespec- trum” aboveOlympusMonsasinPerrieretal.(2006).Thiselimi- natestheriskofsystematicerrorduetotheunwantedpresenceof dustorcloudsinthereferencespectrum,andallowsusingareasof theCCDthatarenotnecessarilythesameasthoseusedtorecord thatspectrum.Otherimprovementsinthemethodincluderevised opticalpropertiesforaerosolsandRayleighscattering,aswellasa shortertime averaging ofnadir spectra(30s vs. 50s)in orderto increasethespatialresolutionofthemeasurements.

Fig. 1 shows the zonally averaged ozone column distribution obtainedfromSPICAMforthreeconsecutiveMartianyearsofob- servation.ComparedtoEarth,thestrongseasonalandspatialvari- abilityofozone(afactorof∼100asopposedto∼3onourplanet) isastrikingfeatureofMarsphotochemistry.Asforthepioneering measurementsoftheMarinermissions(e.g.,Barthetal.,1973),the maximum ozonecolumns measured by SPICAM are near40

μ

^m-

atmandarefoundathighlatitudesinwinter/spring.Bycontrast, polarozone shows columnsusually lessthan 1

μ

^{m-atm insum-}

mer. At mid-to-low latitudes, the ozone annual cycleis of much morelimitedamplitude.SPICAMindicatesvaluesingeneralofthe

order of1

μ

^{m-atm orless, exceptaround Mars aphelionwhen a}

weak maximum(∼3

μ

^{m-atm)isidentified. Thisozonecycleisin}

goodagreementwithHSTandEarth-basedmeasurements(Clancy etal.,1999;Fastetal.,2006),aswellaswiththeobservationsby theMARCIimageronboardMRO(Clancyetal.,2016).

Anindirectmeasurementoftheozonecolumnmayalsobeac- complishedvia theO₂(a¹࢞g)dayglow detected withthe IRchan- nel.ThisairglowisproducedduringozonephotolysisbysolarUV radiation. The O₂(a¹࢞g) molecule, that is an excited state of O₂, can subsequently emit a photon at 1.27

μ

^{m or be quenched by}

CO₂. The latter process dominates at altitudes below 20km and thereforedayglowonlyreflectstheozoneabundanceabove20km.

The first seasonal evolution of the O2(a¹࢞g) emission has been obtainedfromthe SPICAMdata andpresentedby Fedorovaet al.

(2006b). This work waslater extended and resulted ina contin- uous climatology ofthe O₂(a¹࢞g) dayglow for sixMartian years from the end of MY26 to MY32 (Guslyakova et al., 2016). The O₂(a¹࢞g)intensityistraditionallyexpressedinMR(MegaRayleigh) unit,where1R=10⁶photon cm⁻²s⁻¹(4

π

^{ster)−1.Fig.1showsthe}

zonally-averaged emission rate for five Martian years. The good correlation withtheozoneseasonalbehavior isobvious.Maximal values ofthe O₂(a¹࢞g) dayglow reaching 30MR are observedin earlynorthernandsouthernspringsinbothhemispheres.Nearthe equator aspringmaximumof7MRisobservedforallyears.The emissionintensityisminimsalinthesouthernhemisphereinsum- merwithvaluesof1–2MR.Togetherwiththeozonecolumnden- sitydeterminedwiththeUVchannel,theO₂(a¹࢞g)emissionmea- suredwithIRchannelcanprovideconstraintsontheverticaldis- tributionofozonefromnadirmeasurements.

The distribution of ozone in Fig. 1 shows an obvious anti- correlation withthat of water vapor. This inverse relationship is well understood from a theoretical point of view. The strongest sink of oddoxygen (O and O₃) is accomplished by the reactions withthehydrogenradicals(H,OH,andHO₂)producedbythepho- tolysis andoxidation ofwater vapor. The efficiencyofthese pro- cesses issuch thatthe chemicallifetimeofozoneduringthe day is,atlow-to-midlatitudes,usuallyshorterthan1hinthelow at- mosphere. Thistight couplingbetween O₃ andH₂O explainsthe near-zeroozone columnsobserved by SPICAMwhen watervapor sublimesfromthepolarcapsinsummer.Italsoexplainsthetropi- calO₃maximumobservedaroundaphelion,whentheconfinement of H₂O near the surface allows more O₃ to form above the hy- gropause.Inthewintertimehighlatitudes,thelackofwatervapor andlimitedsunlightprevent almostcompletelytheproductionof HOx.In theseconditions, thechemical lifetime ofozone mayin- creaseuptoseveraldays.Ozonecanaccumulateinsidethedrypo- larvortices, andatmospherictransport playsa majorrole. Three- dimensional modelsimulations haveshown(Lefèvreet al., 2004) that wave activityexplainswell theconsiderableday-to-dayvari- ability in polarozone measured by SPICAM (Perrieretal., 2006) andMARCI(Clancyetal.,2015).

Ozone is therefore a sensitive tracer of both transport pro- cesses(atpolarlatitudes)andoftheHOx chemistrythatstabilizes theCO₂ atmosphereofMars(everywhereelse).Withtheir quasi- pole-to-pole coverage, SPICAM measurements have already indi- catedthatheterogeneouschemicalprocessesinvolvingHOxmaybe important to explain quantitatively the ozone amounts retrieved fromthe instrument(Lefèvre etal., 2008). Furtherconstraintson photochemical models are anticipatedfrom theunique capability of SPICAMto retrievesimultaneously the ozoneand watervapor columns,whichwillbepresentedina detailedstudycurrentlyin preparation.

•Ozoneprofilesderivedfromstellarandsolaroccultation Acompilationofnighttime ozoneprofileswasmadewithSPI- CAM UV covering the northern spring equinox (solar longitude Ls = 8° ) to northern winter solstice (Ls = 270°) time frame

(Lebonnoisetal., 2006).Theseobservationsalongwiththeoretical studieshaveallowedcharacterizationofthenighttimeozoneverti- caldistributionatmid-to-lowlatitudes,yieldingacleardistinction between a near-surface layer, typically perceptible below 30km, andan elevated layeronlypresentaroundaphelion.As explained in a numberof previous studies (Lefèvre et al., 2004; Lebonnois etal.,2006),thenear-surfacelayerexistenceispromotedbyultra- violetscreeningbyCO₂ molecules,whichinhibitsHOx production through H₂O photolysis.The elevated layer (> 20kmof latitude) ofozoneappearsessentiallyatnightasO₃israpidlyphotolizedaf- tersunrise.ItsmagnitudeiscontrolledbyHOxconcentrationabove 25km.Forthisreason,seasonalvariationsofthewatersaturation levelexert a strongmodulation oftheelevated layer ofozone in thetropics,forcing O₃ to maximize aroundthenorthern summer solstice when temperatures are lowest and to vanish soon after Ls ∼130. In a recent study (Montmessin andLefèvre, 2013), the analysisoftheozoneprofilesobservedwithSPICAMwasextended tothe polarregions andevidencewasreported fortheexistence ofanozonelayerthat hasbeenobserved repeatedlyduringthree consecutiveMartianyearsinthesouthernpolarnightat40–60km inaltitude,withnocounterpartobservedatthenorthpole.Com- parisons with global climate simulations indicate that this layer forms as a result of the large-scale transport of oxygen-rich air fromsunlitlatitudestothepoles,wheretheoxygenatomsrecom- bine to form ozone during the polar night. However, transport- drivenozoneformationiscounteractedbythedestructionofozone through reactions with hydrogen radicals, whose concentrations varyseasonallyonMars,reflectingtheseasonalvariationsofwater vapor verticaldistribution inthe tropics,asdiscussed above.The observed dichotomy betweenthe ozone layers of the two poles, witha significantly richerlayer inthe southern hemisphere, isa robustindicatorofthecouplingbetweenchemistry,transportand watercycleprocesses.

OzoneprofilescanalsobederivedfromSPICAMUVsolaroccul- tations,whichwereanalyzedforthefirsttimebyMäättänenetal.

(2013).However,theseauthorsonlydiscussedthederivedaerosol andcloud verticaldistribution. Here, wepresent afirst subset of theozoneprofilesobtainedfromsolaroccultationthatweplanto analyzeinthefutureagainstthewatervaporprofilesretrievedsi- multaneouslywiththenearIRchannel.InFig.7,we comparere- sultsforozoneasobtainedfromstellar(lowerpanel)andsolaroc- cultations(upper panel)forthelatitudeband 0–30°S.The plotin thelowerpanelisan extendedversionofthefigurepresentedby Lebonnois et al. (2006). Whereas the stellar occultation pertains toawide rangeoflocaltimesinthenight,solaroccultationonly dealswiththemorning andevening terminatorconditionswhere the chemical regime is in a transient state. The upper panel of Fig.7showsthatasimilarlayeratthesamealtituderangeisob- servedbothatnightandnearterminator.However,thelayeratthe terminatorappearssignificantly weakerthanthenight-timelayer.

Lebonnoisetal.(2006)discussedthedailycycleofozoneandno- ticedthattheozonelayeraround30–45kmaltitudeformsrapidly aftersunset,buildingupuntilsunrise.Sincesolaroccultationsoc- curatsunset/sunrise,thelayerhaseithernotfullyformed(sunset) orhasalreadystartedtowithstandphotolysis(sunrise),explaining theloweramountsobtainedinthecaseofthesolaroccultation.

AjointanalysisoftheSPICAMozoneandwatervaporproducts in solar occultation, the latter beingprovided by the SPICAM IR channel (Maltagliati etal., 2011, 2013), will enableacomprehen- sive assessment ofthe theory governing the ozone chemistry on Mars.

3.3.Dustandclouds

•Nadirmapping

SPICAM Seasonal evolution of ozone, 0-30 S

0 50 100 150 200 250

Solar longitude 0

20 40 60

Altitude (km)

SPICAM Seasonal evolution of ozone, 0-30 S

0 50 100 150 200 250

Solar longitude 0

20 40 60

Altitude (km)

0 2 4 6 8 10 12

x109cm-3

Fig. 7. Seasonal evolution of ozone concentration profiles in the latitude band 0–30 °S. (Upper panel) Solar occultations in the UV. (Lower panel) Stellar occultations in the UV. The color code shows ozone concentrations in 10 ⁹molecules ×cm ^-3. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

The SPICAM UV spectral domain allows the detection of air- bornedustandcloudparticles.Icecloudsanddustbothplayarole intheradiativetransfer buttheir effects areverydifferentin the UV.Icecloudsarebrightfeaturesthatlie aloftagainstadarkand absorbingsurface.Dustontheother handpossessesoptical prop- ertiesthatarecomparabletothoseofthesurface,makingitmore complicatedtodistinguish.Nadir UVdatafromSPICAMUVobser- vationsinnadirmodewasalsousedtostudyMartianaerosols.

Dust abundancesvarygreatlydependingontheseasonandlo- cationon the planet (see e.g. Smith, 2004). During mostpart of the year, dust is present in relatively low-to-moderate quantity andislocatedlowintheatmosphere(below30km).Howeverthe planetundergoes regular duststorms, especially duringsouthern spring and summer, when larger amounts of dust are lifted up intheatmosphereandcanreachaltitudes up to60km.The dust storms season happens typically during the Ls= 180–330° time- frame.NadirUVSPICAMdatahasprovidedan opportunitytoim- provetheknowledgeabouttheMartiandustoptical propertiesin theUVdomain(see thefirststudyinPerrieretal., 2006). Analy- sisoftheSPICAMnadirspectraacquiredduringafewduststorms allowedustoestimate opticalproperties,suchassinglescattering albedo

ω

^{andasymmetryfactorg(Mateshvilietal.,2007).There-}

portedvaluesare

ω

=0.6±0.045,g=0.88±0.04at

λ

=213nm and

ω

=0.64±0.04,g=0.86±0.03at

λ

=300nm.Watercloud opticaldepth distributions duringdifferent seasonswere alsore- trieved (Mateshvili et al., 2006, 2008) and improved in a recent investigation(Willameetal.,2017),which analyzedthecomplete SPICAMdataset available from MY26 to the end ofMY30. Cou- plingtogether theretrievalsonthesefouryearsoffersarelatively

complete overviewthe Martianannualcloud distribution outside thepolarcaps(seeFig.8).

Overall,theicecloudabundancededucedfromSPICAMUVdata showsthesamefeaturesastheonesalreadyreportedinthepast byanumberofauthors(e.g.Clancyetal.,1996;Wolff etal.,1999;

Smithetal.,2004).Inparticular,itshowsveryprominentseasonal variationsandspatialdistribution(Fig.8).Martianiceclouddistri- butionduringthe yearlycycleisclearlyconnectedwiththe Mar- tian atmosphericgeneral circulation and the watercycle. During the northern winter, thick clouds cover the Martian North Pole, formingthenorthpolarhood.Asspringprogresses,cloudsretreat togetherwithseasonalicetotheNorthandonlyafewstayabove the North Pole during summer. The released watervapor rushes towardstheequatorwhereitispickedupbytheascendingbranch oftheHadleycirculation.Therisingwatervaporreachesaltitudes wherelowtemperaturesandpressures createfavorableconditions forcloud condensation.Cloudsformthe so-calledAphelion cloud belt,whichispersistentduringthewholenorthernsummer.Atthe same time, cloud cover grows above the South Pole forming the south polarhood. When the northern fall starts, the northpolar hoodgrows andthesouthpolarhood retreats.Due tothehighly ellipticalorbitofMarsthesouthernsummerismuchwarmerthan thenorthernone.Duringtheseasonofduststorms,cloudsdisap- pearalmosteverywheresurvivingonlyabovetheNorthPole.Only duringsouthernfall rarecloudsstart toformabovehighMartian volcanoesandthecloudcyclestartsagain.

•Verticalprofilesfromsolaroccultations

SPICAMIRhasprovidedlong-termobservationsoftheMartian aerosolsbysolaroccultationtechniquesinthenear-IRrange.Dur- ing thefiveMartianyearsabout880successfulsolaroccultations

Fig. 8. Zonally averaged cloud optical depth derived from SPICAM nadir UV observations and displayed as a function of latitude and solar longitude. The map assembles four consecutive years of retrieval. Note that the regions covered by ice were rejected from the retrieval.

havebeenperformed. SPICAMspectralrangeallows simultaneous observationsof1.43

μ

^mCO2bandfortheatmosphericdensity,the 1.38

μ

^{mabsorptionbandofwatervapor(aspresentedabove),and}

thedistributionofaerosolswithaltitudemeasuringtheopacityof theatmosphereinthespectralrangefrom1to1.7

μ

^m.Inthiscon-

figuration, aerosol extinction profiles at 10 different wavelengths have been obtainedat altitudes from 0 to 90km witha vertical resolution of 2 to 10km depending on the distanceto the limb.

We considertheir seasonalandlatitudinal variations(Fig.9). The interpretationusingMiescatteringtheorywithadoptedrefraction indicesofdustandH₂Oiceallowstheretrievalofparticlesizedis- tributionforr_effvaryingfrom0.1to1.5

μ

^{maswellasnumberden-}

sityprofiles.

A special attention was paid to the summer in the northern hemisphere, when the water vapor supersaturation in the mid- dle atmospherehas beendiscovered (Maltagliati etal., 2011).Si- multaneousanalysisofaerosolextinctionsintheUVandIRrange at differentaltitudes duringthis periodhas enabled the first di- rect detectionof a bimodal distributionof Martiandust particles withcharacteristiceffectiveradiiof0.04–0.07

μ

^{mand0.7–0.8}

μ

^m

(Fedorovaetal., 2014).Thefinermode withaeffectiveradiusbe- tween 0.04and 0.07

μ

^{m anda numberdensities from1cm−3 at}

60km to 1000cm⁻³ at 20km has been detected in both hemi- spheres. In the southern hemisphere the finer mode extends up to 70km,whereas in thenorthern hemisphere it isconfined be- low 30–40km.Thelackofcondensationnucleiisconsistentwith, butcannotfullyexplainthehighwatersupersaturationlevels ob- served between 30 and50km inthe same northern hemisphere dataset(Maltagliatietal.,2011).Theaveragesizeofthefinemode (∼50nm) andthe large number density(up to 1000cm⁻³) most likelycorrespondstoAitkenparticles(r<0.1

μ

^{m).Thismodeisun-}

stableagainstcoagulationandrequiresacontinuoussourceofpar- ticlestobemaintained,inatleastoneorderofmagnitudegreater than estimations for the meteoric flux. The sedimentation time varies from100 to 1000 days (for particles 0.1 and0.01

μ

^m,re-

spectively)at20kmand10–100daysat40km,andtheseparticles may be transportedby the Hadleycell from the northernto the southern hemisphere inthe observed periodof the summer sol- sticeinthenorthernhemisphere.

SPICAMsolar andstellaroccultationsin theUV havebeenre- centlycombinedinto afourMartianYears datasetonthe aerosol

vertical distribution. The first analysis of stellar occultations by Montmessinetal. (2006a) has beencomplemented by an analy- sisofthesolaroccultationdatasetcarriedoutbyMäättänenetal.

(2013) comprising more than four Martian years of data. Fig. 10 showstheinter-annual variationsofthehazetop duringMartian Years27–30andthedetecteddetachedaerosollayersintheatmo- sphere asa function of season. There is a clear trendfor higher hazetopsneartheequatorandduringthesecondhalfoftheyear, drivenbymorevigorousatmosphericcirculationattheselocations andseasons.Thedustyseasonofsouthernsummerismostevident intheoccultationsperformedinMY28 andMY29. MY27wasrel- ativelyaverageintermsofdustinessandSPICAMindicatesamore quiescentsecondhalfoftheyear.InMY30,despitethepoorcover- ageinthetropics,SPICAMwasabletocapturethelowhazetops athigherlatitudes.Overall,thelow springhazetopsinmid-and highlatitudesarefaithfullyreproduced bytheMartiandustcycle, observedinMarsYears27(southonly),29,and30.

Detachedlayers apparent inmost profiles followthebehavior ofthehazetop,withthelowestlayersobservedinmid-andhigh latitudesinthebeginningoftheyear.Aroundtheautumnequinox thelayeraltitudes increaseandremainhighthroughoutthelatter halfof theyear. Duringthe dust storm yearMY28 detached lay- erswereobservedatveryhighaltitudes(red-orangecolorsaround Ls 260°–300°,Fig. 10).The composition oftheselayers cannotbe directly derived, but some hints could be derived from their UV spectralcharacteristicsbyapproachingthewavelengthbehaviorby apowerlaw(Määttänenetal.,2013)

3.4.FromthesearchofSO₂toapreliminarycharacterizationofUV surfacealbedo

Followingthemeasurements ofSO₂ abovethecloudsofVenus by the companion instrument SPICAV onboard Venus Express (Marcqetal., 2011,2013),a similarapproach wasapplied toSPI- CAM data in order to identify SO₂ in the Martian atmosphere andpotentially exploreongoing outgassing,aiminginanycaseat better constrainingSO₂ presence orits absence thereof. Previous work onthe subject fromEarth-based instruments (Nakagawa et al., 2009; Krasnopolsky, 2005) only derived upper limits on the order of 1ppbv on a disk-averaged scale. Preliminary modeling showedthatSPICAMsensitivitytoSO₂ wasontheorder of10 to 100ppbv in cleardust opacity conditions (

τ

< 0.5), but SPICAM

Fig. 9. Seasonal and spatial distribution of the extinction (km-1) vertical profiles at 1094 nm averaged on a grid 30 °of Ls ×5 °of latitude ×6 km of altitude for 4 MY (MY27-31) from orbit 1200 to orbit 12,166.

0 2000 4000 6000 8000 10000

Orbit number -50

0 50

Latitude

MY27 MY28 MY29 MY30

0 100 200 300

Ls -50

0 50

Latitude

0 15 31 47 63 79 95

Fig. 10. (Top panel) Interannual variations of the hazetop defined as the altitude at which the aerosol line-of-sight integrated opacity drops to a value of 1 for the first time starting from the bottom of the profile, and (bottom panel) detached layer altitudes analyzed from stellar (stars) and solar (circles) occultations in the UV. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

wasnonethelessusedtospotlocalenhancementssinceSO₂isex- pected to be locally and temporally variable due to its putative sources(sporadicvolcanism) andsinks(fastphotolysis). However, SO2 mayhave beendetected atsuch levels in its strongestband near210nmwheretheSun’sbrightnesssharplydecreaseswithde- creasingwavelength. The imperfect removal of stray light makes

Fig. 11. The surface radiance factor of Mars (colored lines) in the UV as derived from 3 UV channel observations and a comparison with the radiance factor deduced for Phobos. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

these detections dubious. It is more likely that SPICAM has es- sentiallyconfirmedthepreviouslyderived upperlimits– withthe most recentestimates established by Encrenaz etal. (2011) who obtained0.3ppbvgloballyand2ppbvlocally.

Aninteresting by-productofthisdetectionattemptwasto de- rive spectral UV albedos of the Martiansurface(Fig. 11): thera- diative transfer model adapted fromPerrier et al.(2006) can be usedto simulatereflectancespectratakingintoaccount dustand

Fig. 12. (Black dots) CO 2density measured by stellar occultation at 100 km above the Mars zero datum (areoid) between January 2004 (MY26, Ls = 333 °here shown at Ls = −27 °) and February 2011 (MY29, Ls = 238 °here shown at Ls = 958 °). Only the data obtained below 50 °latitude are shown. (Green curve) A synthetic density ρ=(5 τ −2) x 10 ^–7 directly derived from the visible aerosol opacity τ measured by the Opportunity rover camera ( Lemmon et al., 2004 ). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

O₃ UVopacitiesbutwithgroundalbedosettospectrallyconstant values bracketing the observed reflectance spectrum (typically 0 and0.2).Inthesinglescatteringapproximation(validwheredust opacity islow enough),itcan beshownthat thereisalinearre- lationship betweentheobserved radianceandthe surfacealbedo.

The two simulated, bracketingradiance factors can then be used to derive a spectrum ofthe UV surfacealbedo. Several such de- rived albedos are shownin Fig. 12. Theyare consistent with UV dustalbedoderivedbyPangandAjello(1977)beyond280nm,but differatshorterwavelengthsandexhibitanarrowsignatureinthe 270–290nminterval.Forthesakeofcomparison,thereflectanceof Phobosasdeduced fromthenumerous dedicatedobservationsof theMartiansatelliteisalsodisplayedinFig.11.Itisratherremark- abletofindthatPhobosalbedodepictsaspectralbehaviorakinto theonefoundforMarssurface.

4. Middleatmospheresounding(50–150km)

4.1. Temperatureanddensityprofilingusingstellaroccultationinthe UV

Thehighsensitivityprovidedbytheobservationsofthesetting orrisingofUV brightstars throughtheatmospherewithSPICAM hasprovidedthefirstremotesensingmeasurementsofthedensity and temperaturein the Martianmiddle atmosphere, between 70 and130km(Quemeraisetal.2006;Forgetetal., 2009).Theonly other observations in that altitude range are the in situ density measurements performedbyentryprobesandaerobrakingspace- crafts. CO₂ absorbs stellarradiationbetween120and190nm and allows a high sensitivityfor a wide range ofpressures, between 10⁻⁶ and1Pa.However,atpressurehigherthan0.1Pa(i.e.below 60and70km),ithasbeenfound thatthe retrieveddensitiescan be significantly underestimated as the stellarspectrum gets fully absorbed in its Far Ultraviolet part (short ward of 180nm), and thetemperatures overestimated.No reasonhasbeenfound sofar toexplainthis“warmbias” althoughseveralpotentialcauseshave beenexplored (CO₂ crosssectiondependenceontemperature, at- mosphericrefraction).

The first616 usableprofilesobtainedfora widerangeof lati- tude andlongitudeover alittlemorethan oneMartianyear(MY 27)havebeenanalyzedbyForgetetal.(2009).SinceMY27,many more occultation observations have been performed, although in

Fig. 13. CO 2density measured by stellar occultation at 100 km during MY27 (black dots) compared with predictions from an improved LMD Global Climate Model ( Madeleine et al., 2011, 2012 ; Navarro et al., 2014 ) using a dust climatology com- piled for MY27 (Montabone et al., 2017, submitted for publication). The agreement is much better than with the previous version of the GCM (see Fig. 11 in Forget et al., 2009 ).

manycasesthemeasurements havebeenaffectedbythespurious intensifierspikesdescribedpreviously(seeAppendix).

Fig.12 showsa compilation of the density measurements in- terpolatedat 100kmabove the Marsgeoid (areoid)between the end of Martian Year 26 until mid-Martian Year 29. As analyzed in Forget et al. (2009), atmospheric densities exhibit large sea- sonalfluctuationsmostlyduetovariationsinthedust contentof the lower atmosphere which controls the temperature, andthus theatmosphericscaleheight, below50km.Thedirectcorrelation betweenthe air densityin the upperatmosphere andthe lower atmospheredust loading is illustrated by the green curve in Fig.

12 which is a linear function ofthe visible dust opacity as mea- sured by the camera on the Opportunity rover (Lemmon et al., 2004).TheresponseoftheupperatmosphereofMarstolowerat- mospheric duststorms hasalso been studiedusing SPICAM data andotherdatasetsbyWithersandPratt(2013).

InForget etal. (2009),comparison ofthe observed variations tothe LMDGlobal ClimateModel (GCM)revealedsignificant dis- agreement mostly due to an inappropriate dust climatology, but also dueto missing physicalprocesses in the model. Since then, the GCM has been improved, with a semi-interactive dust cycle torepresentthedustverticaldistributionandtheinclusionofthe radiative effectof watericeclouds (Madeleine etal., 2011, 2012; Navarro etal., 2014). Combined withan improveddust climatol- ogy, this has allowed a much better model-data agreement and thusabetterunderstandingoftheseasonalvariationoftheatmo- sphericdensityobserved bySPICAMasshowninFig. 13.SPICAM densityandpressurevariationmeasurementshavealsobeenused to detect andcharacterize non-migrating tides in the 70–120km regionbyWithersetal.(2011).Theyfoundthatthesignalisdom- inated by wave-2 and wave-3 components andthat they mostly resultfromthediurnalKelvinwaves1and2.

Temperatureprofiles have beenderived fromthe densitypro- files assuming that the atmosphere is in hydrostatic equilibrium (Quémerais etal., 2006; Montmessin et al., 2006b). The SPICAM measurementsrevealedinteresting,butlimitedseasonalvariations at a given pressure level, but large day-to-day and longitudinal fluctuations.Thediurnalcyclecouldnotbe analyzedindetailbe- causemostdatawereobtainedatnighttime,exceptforafewoc- cultations observed around noon during northern winter. There, theaveragedmidday profilewasfound todifferslightlyfromthe correspondingmidnightprofile,withthe observeddifferencesbe- ingconsistentwithpropagatingthermaltidesandvariationsinlo- calsolarheating.

About 6% of the observed mesopause temperatures in MY27 exhibited temperature below the CO₂ frost point, especially

Fig. 14. Average temperature profiles measured by SPICAM (black lines), simulated by the version of the LMD-MGCM described in Forget et al. (2009) (green lines) and by the improved version of the model (red lines, see González-Galindo et al., 2013 ). The observations are averages of 16 profiles obtained between Ls = 90 °–120 °and latitude 17 °S–16 °S (left panel) and of 48 profiles obtained between Ls = 240 °and 270 °and latitude between 30S and 10 N (right panel). The modeled profiles are temporal and geographical averages for the same temporal and geographical intervals. Note that SPICAM temperatures below the 0.1 Pa level are thought to be overestimated. Approximate altitudes are given for reference. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

duringNorthernsummerinthetropics.Theveryhighlevelsofsu- persaturation encountered suggest that CO₂ ice cloudssimilar to theonedirectlyobservedduringthedaybynear-infraredimaging (Montmessinet al., 2007;Määttänen etal., 2010) mayalso form atnight.Montmessinetal.(2006b)analyzedfourexamplesofSPI- CAMstellar occultation observationsof unusual detached aerosol layersaround100km,thoughttobeclouds,andfoundthatineach casesthesimultaneous temperaturemeasurements indicatedsub- freezingtemperatureafewkilometersabove.

SPICAMtemperaturemeasurementshaveprovideda first-hand validationforthe thermospheric models.As detailedin Forgetet al.(2009) andGonzalez-Galindo etal. (2009) forthe LMD GCM, aswell asMcDunnetal.(2010) fortheMarsThermosphereGCM developedattheUniversityofMichigan,itwasfoundthatmodels stronglyoverestimatedtemperatures,i.e.themesopausewasfound attoolowaltitudes.Thisdiscrepancywasanalyzedandattributed totheparameterizationofthecoolingduetoCO₂15

μ

^memissions

inNon-LocalThermalEquilibriumconditionsasthelikelyrespon- sible.Forinstance,theparameterizations usedatLMDincludeda numberofsimplificationsthat affectedthe radiativetransfer, and madeuseofafixed,constantanduniformatomicoxygenconcen- tration.MotivatedbytheSPICAMdata,animprovedparameteriza- tionhasbeendevelopedtoovercomethoselimitations(seeLópez- Valverde et al., 2011 and González-Galindo et al., 2013) and the actualatomicoxygenprofilecalculatedbythemodelisnowtaken intoaccount.Fig.14showstwo averageSPICAMtemperaturepro- filesobtainedatdifferentseasonsandlatitudes,compared tothe average temperatureprofiles predictedby the old version ofthe LMDGCM(andasshowninForgetetal.,2009,Fig.15)andbythe updatedversionofthemodel.Theagreementismuchbetter.

4.2.Ground-statemolecularoxygenO₂

Understanding the distribution of ground-state O₂, a primary photochemicalproductonMars,isimportantinconstrainingpho- tochemical,dynamicalandevolutionarymodelsoftheMartianat- mosphere. For some reason, Sandel et al. (2015) have reported altitudeprofiles of O₂ from SPICAM UV observations of sixstel- laroccultations.Theyinferredthe O2 abundancefromabsorption by O₂ in the Schumann–Runge continuum, which lies shortward of 175nm. Although the absorption cross-section of O₂ is larger

Log

10

(O

2

Mixing Ratio)

-3.5 -3 -2.5 -2 -1.5

Altitude (km)

90 100 110 120 130 140 150

0395 0687 0718 0887 0977 1610 Viking 1 Viking 2

Fig. 15. The circles mark O 2mixing ratios inferred from six UV stellar occultations ( Sandel et al., 2015 for orbits/Ls of 0395/32,7 °, 06 87/6 8.9 °, 0718/78.7 °, 0887/93.5 °, 0977/104.7 °, 1610/195.4 °). Error bars are omitted for clarity; typical values are in the range 10–25%. The numbers in the legend are the Mars Express orbit numbers.

Geometry information about the sampled regions may be found in Table 2 . The tri- angles indicate values from the lower part of the altitude range sampled by Viking.

These profiles suggest variations in O 2mixing ratio in space and time.

than that of CO₂ over much of the wavelength range of inter- est, the low mixing ratio of O₂ relative to CO₂ means that the O₂ absorption signature is at best a small feature in an absorp- tion spectrum dominated by CO₂. This O₂ absorption signature was reliably measured and O₂ abundances inferred over an alti- tude range that varied among occultations, typically from 90 to 130km, which spans much of the region of most active photo- chemistry. CO₂ abundances determined simultaneously from the sameobservationspermit computingtheO2 mixingratios,which arelistedinTable2andplottedinFig.15.Thesemixingratiosare generallylargerbya factorofthreetofourthanvaluesfoundfor

Table 2

Characteristics of the observations used by Sandel et al. (2015) to infer the concentration of molecular oxygen in a fundamental state.

Time Orbit Lat ( °) LST (h) L S( °) O 2vmr ^a,b 2004-05–13/10:34 0395 16.7 21.05 32.7 5.3 ±1.1 2004-08–03/02:49 0687 −38.0 18.67 68.9 5.8 ±1.5 2004-08–11/20:14 0718 −75.4 7.29 72.7 3.2 ±1.1 2004-09–28/04:56 0887 −75.1 3.77 93.5 3.6 ±1.3 2004-10–23/06:22 0977 −16.4 3.74 104.7 3.1 ±0.6 2005–04–18/13:54 1610 37.7 21.99 195.4 3.3 ±1.8

a Volume mixing ratio in units of 10 ⁻³.

bUncertainty reflects variation over altitude, not measurement uncer- tainty.

the lower atmosphere(see e.g. Owen etal., 1977;Hartogh etal., 2010;Mahaffyetal.,2013),butthelowerSPICAMprofilesaresim- ilartothosemeasuredbyViking(NierandMcElroy,1977)atsimi- laraltitudes.Asanon-condensiblespecies,O₂ locatedclosetothe surfaceof Marsisexpected toexhibit seasonal variabilitycaused byenrichment/depletionofnon-condensiblespeciessubsequentto CO₂ freeze/thaw cycle. Our limited dataset does not allow us to conclude on a propagationof this seasonal signatureat thealti- tudeswhereitispossibleforSPICAMtodetectO₂.

4.3. O₂(a¹

g)nightglow

TheO₂(a¹

^{g) nightglowat1.27}

μ

^{misasensitivetracer ofthe}

middle and upper atmosphere circulation. Unlike the O₂(a¹

g) dayglow emission that results from the O₃ photolysis (see also Guslyakova etal., 2014,2016),the O₂(a¹

^{g)nightglow isa prod-} uct oftherecombinationofOatomsformed byCO₂ photolysis at altitudeshigherthan80kmandtransportedtothewinterpoleby theHadleycirculation.ThefirstdetectionsoftheO₂(a¹

^g)night- glow byOMEGA andCRISMshowed thatit istwo orderof mag- nitudelessintense emissionthan theO₂(a¹

g)dayglow (Bertaux et al., 2012; Clancy et al., 2012). In 2010 the vertical profiles of theO₂(a¹

g)nightsideemissionwereobtainedwithSPICAMnear thesouthpoleatlatitudesof82–83°Sfortwosequences ofobser- vations: L_s=111°–120°and L_s= 152°–165°(Fedorova et al.,2012).

The altitudeoftheslantemission maximumobservedbySPICAM variesfrom45kmatLs=111°–120°to38–49km atLs=152–165°.

Onceconvertedintoaverticallyintegratedintensity,theemission attheselatitudesshowsan increasefrom0.22to0.35MR. Those values are consistent with the OMEGA observations reported in Bertaux etal.(2012).From the O₂(a¹

g) nightglowit ispossible to estimate the densityof oxygen atoms at altitudes from50 to 65km,which is found to vary from 1.5×10¹¹ to 2.5×10¹¹cm⁻³. Comparison with the LMDGCM withphotochemistry (Lefèvre et al., 2004,2008) shows that themodel reproduces fairly well the O₂(a¹

g) emission layer observed by SPICAM at night, provided that theintrinsicvertical resolution(> 20kmonthelimb)ofthe instrumentistakenintoaccount.

Since2011SPICAMIRhasbeencarryingoutacampaignatthe northern and southern polar latitudes, where about 120 vertical profiles oftheO₂(a¹

g) nightglowhavebeenobtained.The com- parisonoftheemissiondetectedinbothpolarregionsshowsthat theemissionpeakislocatedat35–42kmabovethenorthpolever- sus40–55kmabovethesouthpole(seeFig.16).Theverticallyin- tegratedemissionrateisabouttwotimessmallerthancalculated by the LMD GCM. Thisdifference mayreflect the current uncer- tainties in the kinetics ofthe production ofO2(a¹

g), or can be duetoaninaccuratedownwardtransportofOatomsby theGCM inthepolarnightregion.

4.4.Nitricoxide(NO)nightglow

In the upper atmosphereof Mars, CO₂ and N₂ moleculesare photodissociatedby solar UV and EUV radiation.O andN atoms are then transported by the mean meridional circulation to the winter pole region. They recombine in the polar night as air parcels subside to form O₂ and NO (nitric oxide) molecules, as revealed by their nightglow emission at1.27

μ

^{m and inthe UV,}

respectively. Therefore, the NO UV nightglow emission in the

δ

(190–240nm) and

γ

^{(225–270nm) bands,which arises from ra-}

diative recombination between O(³P) and N(⁴S) forming excited NO molecules, is a good tracer of the dynamics of the Martian polar night between 30km and 100km. Following the discov- eryby SPICAM ofthe NO nightglow(Bertaux etal., 2005b), Cox etal. (2008) later analyzed 21 orbits of SPICAM containing limb observations of these NO UV emissions. This allowed an exten- sive description of the variability of the NO emission brightness peakandits peakaltitude,mainly dependingon thelatitudeand the season. The upper panel in Fig. 17 illustrates an example of an NO emission limb profile. From the comparison with a one- dimensional chemical-diffusive model, valuesfor the oxygen and nitrogendensityprofiles(middlepanel)andtheeddydiffusionco- efficient(lowerpanel)havebeendetermined(Coxetal.,2008).The downwardfluxofNatomsat100kmvariesbytwoordersofmag- nitude,rangingfrom10⁷to10⁹atomscm⁻²s⁻¹.Oxygenandnitro- gendensityprofilespresentapeaknear50kmand70km.Coxet al.(2008)adoptedan altitudedependenceK(z)=A /N^1/2,where Nisthetotalnumberdensityincm⁻³andAisaconstantbestfit- tingthe observed limb profile.Theyfound A values inthe range 6.4×10⁵ to1.5×10⁶,corresponding toKvaluesaboutanorderof magnitudelessthanothervaluesintheliterature.

Tothe above-mentioned dedicatedlimb dataset wasaddeda largenumberofstellaroccultations,fromwhichtheNOsignalwas extractedfor128occultations(Gagné etal.,2013).Overall,theSPI- CAMdatacoveralmost threeMartianyears fromLs = 44°ofMY 27toL_s=326°ofMY29(Fig.18).Aroundthesouthernwintersol- stice,allpositiveNOdetectionsbetweenLs=60°andLs=120°are atlatitudessouthwardof30°S.ThisisinagreementwiththeLMD GCM calculations, which show a strong NO emission in the de- scendingbranchoftheHadleycirculationabovethesouthernpolar region.Incontrast,aroundthenorthernwintersolstice(Ls=240°–

300°),theSPICAMdetectionsofNOarespreadoveralargelatitude range,from20°Sto70°N,whereastheLMDGCMpredictsamuch weakeremissionneartheequator thanathighnorthernlatitudes.

AnotherNorth-SouthasymmetryisthattheSPICAM“NOpolarsea- son” appearstobemuchshorterforthesouthernhemispherethan forthenorthernhemisphere.Thisisunexpectedsincethesouthern winterislongerthanthe northernwinter,duetoorbiteccentric- ity.Forequinoxconditions(aroundL_s=0°andL_s=180°),SPICAM doesnot detectbrightNO emissionspolewardof60°,contraryto thesimultaneous NO maximacalculated by the LMDGCMabove the poles inboth descending branchesof the Hadleycirculation.

ThesediscrepanciesbetweenSPICAMdataandcurrentmodelcal- culationsoftheNOemissionarestillnotunderstood.

A comprehensive survey of the SPICAM relevant database for NO(i.e.stellaroccultationsandlimbobservations) wasconducted by Stiepen etal. (2015b) to further characterize the factors that control the NO nightglowemission layer. Out of a total of 5000 observations, NO nightglow was detected in 200 of them. They confirmedpreliminary results by Cox et al.(2008) and Gagné et al.(2013)regardingtheemissionpeakaltitude(72±10.4km)and brightness(5± 4.5kR).TheyshowedthatthelargeNOnightglow layer, from 42 to 97km, experiences large brightness variability, rangingfrom0.23to 18.51kR.Stiepen etal.(2015b)showedthat thenumberofdetections increaseswithhighersolaractivity.Yet, thesolar activitydoesnot play a role in theNO nightglow layer