Molecfit: A general tool for telluric absorption correction. I. Method and application to ESO instruments

1.6 MB·English

by  A. Smette

Astronomy&Astrophysicsmanuscriptno.molecfit (cid:13)cESO2015
January29,2015
Molecfit: A general tool for telluric absorption correction(cid:63)
I. Method and application to ESO instruments(cid:63)(cid:63)
A.Smette1,H.Sana1,2,S.Noll3,H.Horst1,4,W.Kausch3,5,S.Kimeswenger6,3,M.Barden7,C.Szyszka3,
A.M.Jones3,A.Gallenne8,J.Vinther9,P.Ballester9,andJ.Taylor9
1 EuropeanSouthernObservatory,Casilla19001,AlonsodeCordova3107Vitacura,Santiago,Chile
e-mail:[email protected]
2 nowatESA/SpaceTelescopeScienceInstitute,3700SanMartinDr,Baltimore,MD21218,UnitedStates
3 InstituteforAstroandParticlePhysics,UniversitätInnsbruck,Technikerstrasse25,6020Innsbruck,Austria
4 Josef-Führer-Straße33,80997München,Germany
5
5 UniversityofVienna,DepartmentofAstrophysics,Türkenschanzstr.17(Sternwarte),A-1180Vienna,Austria
1
6 InstitutodeAstronomía,UniversidadCatólicadelNorte,AvenidaAngamos0610,Antofagasta,Chile
0
7 InternationalGraduateSchoolofScienceandEngineering,TechnischeUniversitätMünchen,Boltzmannstr.17,D-85748Garching
2
beiMünchen,Germany

n 8 UniversidaddeConcepción,Casilla160-C,Concepción,Chile
a 9 EuropeanSouthernObservatory,Karl-Schwarzschild-Strasse2,85748GarchingbeiMünchen,Germany
J
8 Received<date>/Accepted<date>
2

ABSTRACT
]
M
Context.TheinteractionofthelightfromastronomicalobjectswiththeconstituentsoftheEarth’satmosphereleadstotheformation
I oftelluricabsorptionlinesinground-basedcollectedspectra.Correctingfortheselines,mostlyaffectingtheredandinfraredregion
. ofthespectrum,usuallyreliesonobservationsofspecificstarsobtainedcloseintimeandairmasstothesciencetargets,therefore
h
usingpreciousobservingtime.
p
- Aims.Wepresentmolecfit,atoolforcorrectingfortelluricabsorptionlinesbasedonsyntheticmodellingoftheEarth’satmospheric
o transmission.Molecfitisversatileandcanbeusedwithdataobtainedwithvariousground-basedtelescopesandinstruments.
r Methods. Molecfit combines a publicly available radiative transfer code, a molecular line database, atmospheric profiles, and
t
s variouskernelstomodeltheinstrumentlinespreadfunction.Theatmosphericprofilesarecreatedbymergingastandardatmospheric
a profilerepresentativeofagivenobservatory’sclimate,oflocalmeteorologicaldata,andofdynamicallyretrievedaltitudeprofilesfor
[ temperature,pressure,andhumidity.Wediscussthevariousingredientsofthemethod,itsapplicability,anditslimitations.Wealso

showexamplesoftelluriclinecorrectiononspectraobtainedwithasuiteofESOVeryLargeTelescope(VLT)instruments.
1
Results.Comparedtoprevioussimilartools,molecfittakesthebestresultsfortemperature,pressure,andhumidityintheatmo-
v
sphereabovetheobservatoryintoaccount.Asaresult,thestandarddeviationoftheresidualsaftercorrectionofunsaturatedtelluric
9
linesisfrequentlybetterthan2%ofthecontinuum.
3
Conclusions.Molecfitisabletoaccuratelymodelandcorrectfortelluriclinesoverabroadrangeofwavelengthsandspectralreso-
2
lutions.Theaccuracyreachediscomparabletoorbetterthanthetypicalaccuracyachievedusingatelluricstandardstarobservation.
7
Theavailabilityofsuchageneraltoolfortelluricabsorptioncorrectionmayimprovefutureobservationalandanalysingstrategies,as
0
wellasempowerusersofarchivaldata.
.
1
Keywords. radiativetransfer,atmosphericeffects,instrumentation:spectrographs,methods:observational,methods:dataanalysis,
0
technique:spectroscopic
5
1
:
v
1. Introduction targetasitwouldappeariftheinstrumentwaslocatedabovethe
i
X atmosphere.Differentapproacheshavebeenproposed.
Ground-basedobservationsand,inparticular,spectroscopycan
r
abe strongly affected by absorption caused by molecules in the Averycommonmethodistoobserveeitherarapidlyrotating
Earth’s atmosphere. The wavelength ranges involved include, early-type or a G-type star (depending on the specific purpose)
butarenotlimitedto,thenear-andmid-infrared.Itisastandard soonbeforeoraftertheobservationsofthesciencetarget.This
practicetocorrectasmuchaspossibleforsuchtelluricabsorp- telluric star should ideally be located close to the science tar-
tionlinesinordertorecoverthespectrum(orphotometry)ofthe getinthesky.Indeed,earlytypestarsareadvantageousbecause
they are mostly featureless, except for hydrogen lines. Spectral
regionscloseinwavelengthtotheoneofhydrogenlines,onthe
(cid:63) Molecfit is available at http://www.eso.org/pipelines/
otherhand,arethereforebestcorrectedbyG-typestars.Therea-
skytools
(cid:63)(cid:63) Based on observations made with ESO Telescopes at the La sontoobservebothsciencetargetandtelluricstarcloseintime
Silla Paranal Observatory under programme IDs: 60.A-9100, 60.A- is to eliminate differences caused by temporal variation in the
9452,076.C-0129,075.A-0603,079.D-0374,080.D-0526,290.C-5022, atmosphereasmuchasthepossible.Ontheotherhand,theneed
092.D-0024,084.D-0912,085.D-0161,086.D-0066,and088.D-0109 toobservebothobjectsangularlycloseontheskyarisesbecause
Articlenumber,page1of18
A&Aproofs:manuscriptno.molecfit
saturated and unsaturared telluric absorption lines do not scale peded any wide-spread use of such tools. The deviation of the
with the same function of airmass. In addition, the distribution modelled absorption line from the observed one can be caused
of molecules in the atmosphere (mainly water vapour) can de- byerrorsintheradiativetransfermodelling,poorrepresentation
pendonthepointingdirection. of the line spread function, and/or limited accuracy of the me-
Observationsofsuchtelluricstandardstarsrequiretelescope teorologicaldata.Inpractice,therefore,ageneraltoolallowing
timethatcouldotherwisebededicatedtootherscientificobser- theuseofsimulatedatmospheretransmissionspectratocorrect
vations.Thisisanecessarybutcostlypractice,especiallyinthe forEarth’satmospheretelluriclinesinawidevarietyofdatais
coming era of extremely large telescopes. Whenever a limited simplynotavailabletotheastronomicalcommunityatlarge.
timewindowofobservabilityisavailable,orinthecaseofhigh- Here we describe molecfit, a tool for modelling telluric
cadencetimeseries,observationsoftelluricstarsfurtherreduce, lines. Molecfit retrieves the most appropriate atmospheric
sometimes considerably, the amount of time that can be dedi- profile (i.e., the variation in the temperature, pressure, and hu-
catedtothemainsciencetargets. midityasafunctionofaltitude)forthetimeofthegivenscience
The quality of the correction can be variable and depends observations. It uses a state-of-the-art radiative transfer mod-
considerably on the difference in airmass between the science ellingoftheEarth’satmosphere,togetherwithacomprehensive
and telluric star observations, because in the best conditions, a databaseofmolecularparameters.Finally,itallowsforfittingan
difference of 1% in airmass already introduces a 1% change in analyticallinespreadfunction,aswellasusingauser-provided
theopticaldepthofunsaturatedtelluriclines.Examplesofother numerical kernel. Molecfit can be used from 0.3 to 30 µm
problemsaffectingobservationsoftelluricstarsarethat(i)suit- (ormore)andcanbeconfiguredtoapplyondatacollectedbya
abletelluricstarsthatarecloseenoughtothesciencetargetmay wide-rangeofopticalandinfraredinstrumentsfrommostobser-
notbeavailable;(ii)atmosphericconditionsmaychangebythe vatories,makingitveryversatile.
endofthescienceexposures;(iii)thesignal-to-noiseratioofthe
The initial development of molecfit was carried out
obtainedtelluricspectrummaybeinsufficient,eitherincompari-
mostly as a set of IDL routines used as drivers for the Refer-
sonwiththequalityofthescienceobservationsorwithrespectto
ence Forward Model2 (hereafter, RFM) radiative transfer code.
one’sscientificaims;(iv)post-observationanalysisofthespec-
The initial goal was to measure the amount of precipitable wa-
trumofthetelluricstarrevealsthatitdoesnothavetheexpected
tervapourtowardszenith(hereafter,PWV)usingthe≈19.5µm
spectral type or characteristics. Critically, telluric star observa-
emissionlinewithVISIR,themid-infraredspectrographandim-
tions may be missing as a result of inclement weather, instru-
agerattheVLT(Smetteetal.2008).Thecodewasthenslightly
mentalissues,humanerrors,and/orincompletecalibrationplan.
modified to measure the amount of PWV using sky emission
The use of archival data provides a particularly illustrative ex-
linespectraobtainedwiththecryogenichigh-resolutioninfrared
amplesinceobservationsofatelluricstarmaybemissingornot
echellespectrographCRIRES,sincethe5.038to5.063µmrange
appropriateforthesciencegoalpursuedbythearchiveuserasit
ispurelydominatedbywatervapour3.
maysignificantlydifferfromtheoneoriginallyintended.
The good quality of the fits indicated that generalising the
Alternatively, the emergence of freely available radiative
codetoabsorptionlinesandotherspectralrangeswaspromising
transfercodesforatmosphericresearchprovidesthepossibility
andledtoawell-developedIDLprototype(Smetteetal.2010).
of using synthetic transmission spectra instead of telluric stan-
ItsperformancesweresimilartotheonecreatedbySeifahrtetal.
dard stars. One of the first articles to demonstrate the ability
(2010)thatonlyfocusedonCRIRESspectra.Thisprototypewas
of this approach is Bailey et al. (2007), who used the radiative
thenportedinarobustwaytoCPL4 andfurtherdevelopedand
transfer model SMART (Meadows & Crisp 1996) for that pur-
optimisedaspartoftheAustrianin-kindcontributionforjoining
pose.Seifahrtetal.(2010)refinedthemethodincorporatingthe
ESO. In the process, the LNFL/LBLRTM code (Clough et al.
radiative transfer code LBLRTM (Clough et al. 2005), the line
2005)waschoseninsteadofRFM,becauseit(i)wasmorefre-
databaseHITRAN(Rothmanetal.2009),acombinationofme-
quentlyupdatedatthetimeofthecodeselection,(ii)isusedby
teorologicaldatafromvarioussourcestoachievesynthetictrans-
a large community, (iii) incorporates a code-optimised and up-
mission curves, and a model of the line spread function. These
to-date line list, and (iv) comprises broader functionality (e.g.,
were then used to successfully perform telluric absorption cor-
UV/optical ozone absorption, which was needed in a parallel
rectionsofCRIRES(Käufletal.2004)spectrauptoanaccuracy
project (Noll et al. 2012)) and performs significantly faster for
ofabout2percent.ThecodeLBLRTMisalsousedbyHusser
most spectral ranges for similar or better precision. This series
&Ulbrich(2013)whoincorporateditintotheirowncodetofit
of papers describes the results of these efforts. In the present
anentireinputspectrum.
A different approach is used by Cotton et al. (2014). Since work,wedescribethekeyingredientsofmolecfit,itsapplica-
bility,andlimitations,andweshowavarietyofapplicationsfor
they investigate solar system planetary atmospheres containing
ESO/VLTdatacoveringarangeofspectralresolvingpowerand
similar molecules to those in the Earth’s atmosphere, they de-
wavelengthregions.Inaseparatepaper,(Kauschetal.2014,Pa-
rive the state of the latter by fitting spectral features of telluric
perII),wehavesystematicallyinvestigatedthequalityofthetel-
standardstarobservations.Gardinietal.(2013)usedtheradia-
luriclineabsorptioncorrectionbyapplyingittothelargesample
tivetransfermodel1developedfortheSCIAMACHYinstrument
of archival VLT/X-shooter near-infrared data, and compared it
onboard the ENVISAT satellite to fit water features visible in
spectratakenattheSierraNevadaObservatory.Inaddition,they
usedittodeterminetheamountofprecipitablewatervapourin
2 http://www.atm.ox.ac.uk/RFM/
theEarth’satmosphereatthetimeofobservation.
3 These VISIR and CRIRES measurements, together with UVES
Unfortunately, the limited quality of these corrections, the and X-shooter measurements obtained using a different method, are
specialised design and/or restricted availability has so far im- available at http://www.eso.org/observing/dfo/quality/
GENERAL/PWV/HEALTH/trend_report_ambient_PWV_closeup_
1 RTMispartoftheSCIATRAN2.2softwarepackage,developedby HC.html.
InstituteofRemoteSensing/InstituteofEnvironmentalPhysicsofthe 4 The Common Pipeline Library http://www.eso.org/sci/
UniversityofBremen,Germany software/cpl/documentation.htmlbasedonANSIC.
Articlenumber,page2of18
A.Smetteetal.:Molecfit:Ageneraltoolfortelluricabsorptioncorrection
withthestandardmethodperformedwiththeIRAFtelluric5 Table1.Specieswithaminorline-absorptioncontribution.Thespecific
spectraarecalculatedwiththeradiativetransfercodeLBLRTM(resolu-
task.
tionR∼10000)withoutcontinuatoextractonlythelinecontribution.
This paper is organised as follows. Section 2 describes the
ThewavelengthrangesarealsoshowninFig.1.
most important aspects regarding the absorption lines formed
in the Earth’s atmosphere. Section 3 provides a description of
themolecfitpackage.Afewcharacteristiccasesaredescribed Species Numberin Wavelengthrange Optical
in Sect. 4. Limitations to the use of the tool are summarised Fig.1 [µm] depth[%]
in Sect. 5. Examples of successful applications of molecfit
on spectra obtained with several ESO instruments are given NO 1 5.153–5.556 <∼0.3
in Sect. 6. The most important aspects of molecfit are sum- HNO3 2 5.743–5.983 <∼1
marised in the conclusion. The technical aspects of molecfit 7.404–7.825 <∼2
and additional user guidance, including a description of the 10.890–11.694 <∼1
GraphicalUserInterface(GUI)thatallowsfordatainteraction, 20.452–23.478 <∼3
are provided in the User Manual (Noll et al. 2013), (hereafter, COF2 3 5.103–5.187 <∼0.01
UM). 7.935–8.157 <∼0.01
H2O2 4 >19.705 <∼0.3
HCN 5 2.959–3.092 <∼1
6.801–7.370 <∼1
2. AbsorptionarisingintheEarth’satmosphere 12.819–15.467 <∼4
NH3 6 5.755–6.530 <∼0.01
The Earth’s atmosphere consists of about 78% of N2, 21% of 8.904–12.571 <∼1
Oth2e,m1o%lecoufleAsra,nadndaersoevsoelrsalafftreacctestghaeselisghatntdraaveerlolisnoglst.hrEoaucghhoinf NO2 7 63..045112––63..348837 <∼<∼04.1
differentwavelengthregimesbyabsorptionandscattering.Fig- N2 8 3.966–4.650 <∼2
ure 1 shows a model transmission curve derived with the sky C2H2 9 12.957–14.523 <∼0.1
model developed in a parallel project (Noll et al. 2012). The C2H6 10 3.332–3.364 <∼2
main absorption regions of the eight major contributing molec- SO2 11 7.243–7.455 <∼0.01
ular species O , O , H O, CO, CO , CH , OCS, and N O are
2 3 2 2 4 2
marked.Thankstothecomplexityofthemolecules,severalro-
vibrationalbandsareclearlyvisible.
3. Descriptionofthemolecfittelluriccorrection
Theopticalregimeisdominatedbybroadabsorptionbands
tool
fromozone(Hugginsbandsshortwardsofλ ∼ 0.4µm,Huggins
& Huggins (1890); Chappuis bands at ∼ 0.5 < λ/µm < 0.7, The molecfit package allows one to simulate and fit tropo-
Chappuis (1880)), narrow oxygen bands (the prominent A, B sphericandstratosphericemissionorabsorptiontelluriclinesaf-
and the weaker γ bands at ∼ 0.759 < λ/µm < 0.772, 0.686 < fectinguser-selectedregion(s)ofanobserved(usually,science)
λ/µm < 0.695, and 0.628 < λ/µm < 0.634, respectively), and spectrum.Itusesparametersprovidedinaconfigurationfileand,
somecomparablyweakwatervapourfeatures.Thelatterbecome if present, the FITS header of the spectrum. It then returns a
dominantintheentireinfraredregimelongwardsofthe I band. transmission spectrum for the whole wavelength range of the
Inadditioninthe JHK regime,thetransmissionissignificantly spectrum.
affectedbyCO2,CH4,andO2.BetweentheKbandand∼18µm Molecfitincludesthefollowingingredients:
absorptionsaremainlycausedbyH OandCO ,withsomecon-
2 2
tributionfromN2O,CH4,andminorabsorptionfeaturesbyCO 1. an automatic retrieval of two atmospheric profiles from the
andOCS.Intheregimefrom18to30µmonlywatervapouris Global Data Assimilation System (GDAS) web site. These
dominant.Overthewholewavelengthrange,thegaseslistedin profiles bracket the time of the observation and can be re-
Table 1 lead to absorption features with optical depth reaching trieved for any observatory in the world. These profiles are
upto5%atthegivenspectraresolvingpower(R∼10000). linearly interpolated to match the time of the observations
The Earth’s atmosphere is highly variable on several time andmergedwithastandardprofileandlocalmeteorological
scalesintemperature,pressure,andchemicalcomposition.The datatocreateaninputprofilefortheradiativetransfercode;
most variable relevant molecule is water vapour, which is di- 2. aradiativetransfercodethatsimulatesatmosphericemission
rectly related to the actual weather conditions at the time of andtransmissionspectra;
observations. However, seasonal, daily, and sometimes hourly 3. amolecularspectroscopicdatabase;
variations in the volume mixing ratio of various gases have 4. asimplegrey-bodycalculatorthatsimulatesthecontribution
been measured, in particular, for CO2 (Thoning et al. 1989), ofthecontinuumfromtheinstrumentandtelescope;
CH4 (Schneising et al. 2009), and O3 (Ramanathan & Ramana 5. achoiceofpossiblelinespreadfunctions(alsoreferredtoas
Murthy1953).Inaddition,followingtheWorldMeteorological kernels);
Organization6,thefractionofCO2hasincreasedby10−15,CH4 6. an automatised fitting algorithm that adjusts the continuum
by∼10,andN2Oby∼8since1985(WorldMeteorologicalOr- spectrum, wavelength calibration, spectral resolution, and
ganization2012)leadingtoglobalwarming.Dailycolumnsfor column density of each of the relevant molecules, with the
a number of molecules (O3, N2O, CH2, etc.) over most of the possibility of excluding spectral regions expressed either in
Earthsurfacecanbeobtainedfromhttp://www.temis.nl/. pixelsorinwavelengthregions.
5 http://iraf.net/irafhelp.php?val=telluric&help=Help+ Inthefollowing,wedescribeimportantaspectsofthesecom-
Page ponentsinmoredetail.Wealsodiscussspecificchoicesmadefor
6 http://www.wmo.int/gaw theESOParanalobservatory.
Articlenumber,page3of18
A&Aproofs:manuscriptno.molecfit
HO HO HO
Chappuis ozone absorption bands 2 2 2
n 1
o
ssi 0.8 HO
Transmi 000...246 O3 (Huggins bands) (γ bOa2nd)2 O2 (B band)
0
O (A band)
2
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
wavelength [µm]
HO HO HO HO
2 2 2 2
1
n
sio 0.8
mis 0.6 CH4
s 0.4 CO
an 0.2 O 2
Tr 2
0
CO2 CO2 CH4
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6
wavelength [µm]
NO O HO
2 3 HO NO 2
2 2 OCS
1
n
sio 0.8
mis 0.6
s 0.4
Tran 0.2 N2O
0
O
CO2 (5) (10) (7) CH4 (8) CO2 CO 3 CO2
2.6 2.7 2.8 2.9 3 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5
wavelength [µm]
(3) HO (11) (3) (9)
2 CO O CO CO
2 3 2 2 O
1 3
n
missio 00..68 (7) N2O
s 0.4
n
a 0.2
Tr
0
(1) (2) (6) (5) (2) CH (6) (2) single H2O lines (5)
4
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15
wavelength [µm]
CO HO dominated regime
O 2 2
1 3
n
missio 00..68 N2O
s 0.4
n
a 0.2
Tr
0
single H2O lines (2) (4)
(5)
15 15.5 16 16.5 17 17.5 18 18.5 19 19.5 20 20.5 21 21.5 22 22.5 23 23.5 24 24.5 25 25.5 26 26.5 27 27.5 28 28.5 29 29.5 30
wavelength [µm]
Fig.1.Syntheticabsorptionspectrumoftheskybetween0.3and30µmcalculatedwithLBLRTM(resolutionR∼10000)usingtheannualmean
profileforCerroParanal(Nolletal.2012).TheeightmainmoleculesO ,O ,H O,CO,CO ,CH ,OCS,andN Ocontributemorethan5%tothe
2 3 2 2 4 2
absorptioninsomewavelengthregimes.Theredregionsmarktherangeswheretheymainlyaffectthetransmission,minorcontributionsofthese
moleculesarenotshown.Thegreenregionsdenoteminorcontributions(seeTable1)fromthefollowingmolecules:(1)NO,(2)HNO ,(3)COF ,
3 2
(4)H O ,(5)HCN,(6)NH ,(7)NO ,(8)N ,(9)C H ,(10),C H ,and(11)SO .
2 2 3 2 2 2 2 2 6 2
Articlenumber,page4of18
A.Smetteetal.:Molecfit:Ageneraltoolfortelluricabsorptioncorrection
3.1. Atmosphericprofile bymolecfitcanbeeasilymodifiedtoanotherexistingorman-
uallycreatedprofile.
Thereferenceatmosphereisthenmergedwiththemodelled
GDAS8 3-D data provided by the National Oceanic and At-
110000
mosphericAdministration9(NOAA).Thismodelcontainstime-
dependentinformationonthetemperature,pressure,relativehu-
) midity, wind speed, and direction up to an altitude of ∼ 26km
m ona3hbasis.Owingtothegeographical1◦ ×1◦ grid,wecre-
k
( ated a set of profiles by means of an interpolation of the four

ht 1100 gridpointsclosesttoCerroParanaltoobtainlocalinformation.
g
ei The GDAS profile derived for a specific observation is then by
H defaultalinearinterpolationofthetworesultingprofilesclosest
in time to the observation. The molecfit package contains all
the available GDAS profiles for Cerro Paranal since December
11
1,2004at00:00UTandthereleasedateofthepackage.Inad-
1100−−44 1100−−22 110000 110022 110044 dition,ESOwillregularlymaintainsuchadatabaseforParanal.
Volume mixing ratio (ppmv) Forobservationsobtainedlaterorfromanothersite,molecfit
automaticallyretrievestheappropriateGDASfiles.
11..00 It may still be possible that no GDAS profiles are avail-
00..99 able for the time of observations. In the case of Cerro Paranal,
y
t molecfitthenusethebestmatchingtwo-monthaverage(Noll
nsi 00..88 etal.2012).Alternatively,anASCIIprofilewiththesameformat
e asaGDASprofilecanalsobeprovidedbytheuser.
t
n 00..77
i Afurtheroptionalrefinementofthefinalprofileisachieved
e
v 00..66 by incorporating on-site measurements obtained by the Astro-
ati nomicalSiteMonitor(ASM)asprovidedintheFITSheaderor
el 00..55 through a parameter file. The influence of the measured values
R
at the observatory altitude is gradually decreased up to a con-
00..44
figurableuppermixingheight,setbydefaultto∼ 5km,where,
forParanal,thewinddirectionchangesbyanaverageof∼ 90◦
55..004400 55..004455 55..005500
onaverage,asrevealedbyananalysisoftheGDASwinddata.
Wavelength (µm)
Thus,itcanbeassumedthatatthisaltitudetheinfluenceofthe
local environment (as determined from the ASM data) has di-
Fig. 2. Top: Two examples of the distribution of the water-vapour
minished. More information on the merging procedure can be
volume-mixingratioasafunctionofaltitude.Theblacksolidlinecor-
foundintheUM.
respondstoavolume-mixingratioof0ppmvupto3.1km(Paranalal-
titude is 2.6km), while the red dashed line corresponds to a constant Water vapour is the molecule showing the most variations
relativehumidityuntil3.1km. Thetotalamountofprecipitablewater becauseitstotalamount(PWV)anditsaltitudedistributioncan
vapour,2mm,isidenticalforbothprofiles.Bottom:sampleofthecor- changesignificantlyandonashorttimescale.Inparticular,the
responding transmission spectra at R = 100000. The higher amount largestamountofPWVmaynotbelocatedinthelowestlayers
ofwatervapourintheloweratmosphericlayersabovetheobservatory oftheatmosphere.Figure2illustratestheimportanceofhaving
leadstobothdeepercoresandstrongercontinuumabsorption. adecentprofileforwatervapour.
ThetotalamountofPWVcanoptionallybesetifthevalue
An atmospheric profile typically describes the temperature can be retrieved independently (for example, from radiome-
T, pressure P, and volume mixing ratio x of several molecular ter measurements, see Kerber et al. (2012)). In this case,
species as a function of altitude h for a given location. These the merged profile composed by the reference atmosphere, the
parametershaveadirectimpactontheshapeandstrengthofall GDAS data, and the local meteo information will be scaled to
the telluric lines. For molecfit we use a merged profile based the requested PWV. Alternatively, the atmospheric profile pro-
on a reference atmospheric profile, modelled 3-D data, and on- videdbytheradiometer–soonretrievablefromtheESOarchive
sitemeteorologicalmeasurements. –canbeuseddirectly.
Thereferenceatmosphereprofileisbydefaulttheequatorial Figures3illustratestheeffectofthevariationoftheground
profilederivedbyJ.Remedios7.Itcontainsabundancesofsev- temperatureandpressureontheH Otelluricspectrum.Depend-
2
eralmolecularspeciesuptoanaltitudeof120km,dividedinto ing on line parameters, different lines behave differently with
121 altitude levels. Owing to the low latitude of Cerro Paranal temperatureorpressure.
(Lat:-24.6◦),theequatorialprofileisbettersuited(AnuDudhia,
private communication) than the tropical or standard profiles,
3.2. Radiativetransfercode
which are also available in the molecfit package. We note here
thatthemolecularcompositionhaschangedsincetheprofilewas Theresultingmergedatmosphericprofileisusedasinputinthe
createdin2001.Inparticular,thecontentofthegreenhouse-gas radiative transfer code LNFL/LBLRTM (Clough et al. 2005).
carbon-dioxide content has increased by ∼ 6% (World Meteo- Theversionusedatthetimeofthefirstreleaseofthemolecfit
rological Organization 2012). Molecfit allows the user to ei- packageisthelatestversionatthattime,V12.2.Thiscodepack-
therfitormanuallyadjustthemolecularcontenttotakethisin- ageiswidelyusedinatmosphericsciencesandispubliclyavail-
creaseintoaccount.Alternatively,thereferenceatmosphereused
8 http://www.ready.noaa.gov/gdas1.php
7 http://www.atm.ox.ac.uk/RFM/atm/ 9 https://ready.arl.noaa.gov/gdas1.php
Articlenumber,page5of18
A&Aproofs:manuscriptno.molecfit
Table 2. List of molecules as provided by the aer line parameter
database.
# Molecule Chemicalname Instandard GUI
profile?
1 H O Water X f
2
2 CO Carbondioxide X f
2
3 O Ozone X f
3
4 N O Nitrousoxide X f
2
5 CO Carbonmonoxide X f
6 CH Methane X f
4
7 O Oxygen X f
2
8 NO Nitricoxide X c
9 SO Sulfurdioxide X c
2
10 NO Nitrogendioxide X c
2
11 NH Ammonia X c
3
12 HNO Nitricacid X c
3
13 OH Hydroxyl
14 HF Hydrogenfluoride
15 HCl Hydrogenchloride
16 HBr Hydrobromicacid
17 HI Hydrogeniodide
18 ClO Chlorinemonoxide X c
19 OCS Carbonylsulfide X c
20 H CO Formaldehyde
2
21 HOCl Hypochlorousacid X c
22 N Nitrogen X c
2
23 HCN Hydrogencyanide X c
24 CH Cl Chloromethane
3
25 H O Hydrogenperoxide X c
2 2
26 C H Acetylene X c
2 2
27 C H Ethane X c
2 6
28 PH Phosphine
3
29 COF Carbonylfluoride X c
Fig.3. EffectofgroundtemperatureorpressurevariationontheH O 2
2 30 SF Sulfurhexafluoride X c
telluric spectrum for two different spectral ranges. Each graph shows 6
31 H S Hydrogensulfide
theratiobetweentelluricspectracomputedwiththeindicatedground 2
32 HCOOH Formicacid
temperatureorpressure.Top: variationof1◦Cand10◦C.Theregion
33 HO Hydroperoxyl
closeto3.68µmshowsoppositebehaviours.Bottom:variationof1mb 2
and10mb. 34 O Oxygen
35 ClONO Chlorinenitrate X c
2
36 NO+ Nitrosonium
37 HOBr Hypobromousacid
able.Itcalculatesradianceandtransmissionspectrawithaspec-
38 C H Ethylene
tralresolvingpowerofaboutfourmillion.Moreinformationcan 2 4
39 CH OH Methanol
befoundatthewebsite10. 3
40 BrO BromineMonoxide
41 C H Propane
3 8
42 C N Cyanogen
3.3. Databaseofmolecularparameters 2 2
In addition to the profile, the radiative transfer code re-
Notes. Col.1:referencenumberofthemoleculeasdefinedintheHI-
quires a line database as input. The first release of molecfit TRAN 2008 database; Col. 2: its chemical formula, followed by its
uses the database aer_v_3.2, which is delivered with the chemicalname.Col.3indicatesifanatmosphericprofileofthemolec-
LNFL/LBLRTMcodepackage:itisbasedonanupdatedversion ularvolumemixingratioisavailableintheequatorialstandardprofile
of the HITRAN 2008 database (Rothman et al. 2009) and con- asincludedin molecfit.Lastcolumn:functionalityavailablethrough
tains information on more than 2.7 million spectral lines of 42 theGUI:f thecorrespondingcolumndensitycanbefitted,cthecorre-
molecular species (see Table 2). The molecfit line database spondingtransmissionspectrumisonlycalculatedusingtheuserpro-
videdabundance,relativetotheonefixedinthereferenceatmosphere.
canbeeasilyupdatedwhenanewversionbecomesavailable.
Only molecules for which an atmospheric volume mixing
ratio is available in the standard atmosphere profile provided
with molecfit can be used by default. These molecules are
3.4. Fittingalgorithm
identifiedinTable2.
Molecfit makes use of the C version of the least-squares fit-
tinglibraryMPFIT(Markwardt2009)basedontheFORTRAN
10 http://rtweb.aer.com/lblrtm_frame.html fittingroutineMINPACK-1(Moré1978).ItusestheLevenberg-
Articlenumber,page6of18
A.Smetteetal.:Molecfit:Ageneraltoolfortelluricabsorptioncorrection
Marquardt technique to solve the least-squares problem lead- 3.8. Wavelengthcalibration
ing to a fast convergence even for the complicated functions
Thesciencespectratoanalyseareexpectedtohaveanaccurate
involvedinmolecfit,whilebeingnumericallyrobustandself-
wavelengthcalibration.However,smallerrorscanaffectthede-
containedandallowingupperandlowerlimitsorparameterstied
terminationofthelinespreadfunctionandcolumndensitiesneg-
toeachotherthatcanbeindividuallyset.
atively. Therefore in each inclusion region, the synthetic spec-
trumbeingfittedcanbeadjustedtomatchtheinputspectrum.A
3.5. Inclusionandexclusionregions Chebyshevpolynomialofdegreen isusedforthispurpose:
w
Aspectrumshowsanumberoffeaturesthatcanbecharacterised (cid:88)nw
inthefollowingway: λ= bt, (2)
i i
i=0
– intrinsicfeaturesfromthescienceobject,
– telluricfeatures, where
– isnptercintrsoicgrfaepahtuorresthferodmetetchteori.nstrument caused either by the t = λ1 ffoorrii==10 (3)
i  2λ(cid:48)t −t fori≥2
Intrinsic features either from the science object or from the in- i−1 i−2
strumentcaninterferewiththefittingprocess.
wherethewavelengthrangeisnormalisedsothatλ(cid:48)rangesfrom
Therefore,molecfitallowsoneto:
−1to1overtheentirespectralrangeofthespectrumor,inthe
caseofmulti-chipinstruments,overtherangecoveredbyanin-
– select inclusion regions, also called fitting ranges, whose
dividualchip.
mainpurposesaretoallow molecfittodeterminethecol-
Theconversionofthewavelengthgridtoafixedintervalre-
umn density of the relevant molecules and the line spread sultsincoefficientsb independentofthewavelengthrangeand
function (or kernel). Such regions must therefore show a stepsizeoftheinputispectrum.Forb = 1andallotherb = 0,
goodenoughcontinuumlevelforthefittingaccuracyandin- 1 i
themodelspectrumremainsunchanged.Thisisthedefaultcon-
cludetelluricabsorptionlinesfordeterminingthelinespread figurationfortheinitialcoefficients.
function.Oneorseveralsuchregionsmustbedefinedsothat
On the other hand, in some cases, one wishes to use the
at least some telluric lines from each molecule to be cor-
telluric lines as references to improve an unsatisfactory initial
rectedarecovered.
wavelengthcalibration.ThisisthecaseforCRIRESinparticu-
– selectexclusionregionswithintheinclusionregionsthatare
lar. For such a purpose, molecfit actually outputs the inverse
affectedbythepresenceofintrinsicfeaturesfromthescience
ofthewavelengthcalibrationdescribedinEq.2,inthesensethat
objects;
the input spectrum is then adjusted to match the synthetic one.
– maskpixelsintheinclusionregionsthatareaffectedpurely Suchwavelengthcalibrationisonlymeaningfulwhentheinclu-
byinstrumentalordetectoreffects. sion range covers the whole spectrum and is sampled well by
telluriclines.
DetailscanbefoundintheUM.
3.9. Linespreadfunction
3.6. Continuumcontribution
Themodelspectrumisconvolvedwithuptothreedifferentpro-
The model spectrum, Fout, is scaled by a polynomial of degree filestodeterminetheshapeofthelinespreadfunction.Theyare:
n :
c
1. asimpleboxcar
(cid:88)nc
F (λ)= F (λ) aλi. (1) (cid:40)
out in i 1 for −w /2≤λ≤w /2
i=0 Fbox(λ)= 0 forλ<−bowx /2orλ>bowx /2 , (4)
box box
Fora =1andallothera =0,themodelspectrumremains
0 i which is adapted to the pixel scale and normalised to an
unchanged.Thisisthedefaultconfigurationfortheinitialcoef-
integral of 1. This kernel is particularly useful for objects
ficients. The continuum correction is carried out independently
fullycoveringtheentranceslit;and
for each fitting range. A fitting range (or the full spectrum) is
splitfurtherifitisdistributedovermorethanonechip.
2. aGaussian
If the input spectrum is an emission line spectrum, an op-
tional flux conversion can be carried out in order to match the 1 (cid:32) λ2 (cid:33)
unit of the synthetic spectrum. Further details are available in Fgauss(λ)= σ√2πexp − 2σ2 (5)
theUM.
centredon0,where
3.7. Telescopebackground
w
σ= √gauss . (6)
For sky emission modelling, the telescope background is as- 2 2ln2
sumed to be a grey body (black body times emissivity). The
parameters of this grey body correspond to the telescope main Thefullwidthathalfmaximum(FWHM),w ,isgivenin
gauss
mirror temperature and an effective emissivity of the telescope pixels.
andinstrument.
Articlenumber,page7of18
A&Aproofs:manuscriptno.molecfit
3. aLorentzian Molecfit allows the user to adjust this quantity based on
the observed telluric lines, and we note that the GUI only al-
1 λ
F (λ)= (7) lows one to fit the column density for the molecules indicated
lorentz π λ2+(w /2)2
lorentz inthelastcolumnofTable2.Thisadjustmentisdonebymulti-
plying the overall atmospheric profile by a single constant. For
centred on 0, where w is the FWHM in pixels. Com-
lorentz
most molecules, the impact is actually limited to the lowest at-
pared to a Gaussian, the Lorentzian approaches the 0-level
mosphericlayers.
fluxsignificantlymoreslowly.
A kernel consisting of all three components is usually too
3.11. Supportedinstruments
complex to be constrained by a fit. To avoid unrealistic best-fit
FWHM, the user should reduce the number of degrees of free- One-dimensional (1-D) spectra of any visible, near-, or mid-
dombyfixingindividualfitcomponents.AFWHMofzeropixel infrared instrument could in principle be used by molecfit.
corresponds to a unity convolution; in other words, it does not DetailsontherequiredinformationcanbeobtainedintheUser
changetheinputspectrum. Manual (Noll et al. 2013). Although the tool has been tailored
Alternatively,theuserhasthepossibilityofchoosingasin- to ESO-Paranal instruments – in particular, the GDAS profiles
gle Voigt profile kernel directly, which is then calculated by areavailableasatarballfileonlyforthissite–therearenospe-
anapproximateformulathattakestheFWHMofGaussianand cificlimitationseitherinthetermsofformatoringeographical
Lorentzian as inputs. The user can also select a kernel whose locationforaground-basedobservatory.
size in pixels linearly increases with wavelength. It is suitable
foraninstrumentalsetupwithnearlyconstantresolvingpower,
fixed wavelength step size, and a wide wavelength range, such 3.12. Inputsandoutputs
ascross-dispersedechellespectrographs.
Molecfitaccepts1-DspectraprovidedasASCIItables,FITS
Finally,auser-definedkernelcanbeprovidedintheformof
tables,andimages.TheFITStablescanhaveseveralextensions
an ASCII table of fixed kernel elements. This option overrules correspondingtodifferentchips(e.g.,CRIRES).Requiredkey-
theparameterisedkerneldiscussedabove.Aninappropriateker-
words are either taken from the FITS header or directly from
nel can cause deviations in the determination of molecular col-
theparameterfile.Finally,ASCIIorFITSfilesforinclusionand
umndensitiesofmorethan10%.
exclusionregionscanbeprovided.
Molecfit fits the synthetic spectrum in the inclusion re-
3.10. Molecularvolume-mixingratio gionsalone.Theseregionscancoverjustasmallfractionofthe
wavelength range or all of it. It returns the best-fit model only
The integrated volume mixing ratio, X(h ), of a molecule for a
0 fortheinclusionregionsinbothanASCIIandaFITStable,and
column of air from the observer at an altitude h to the top of
0 the best-fitparameters and molecularcolumn densities (includ-
atmospherecanbederivedby
ingthePWV)inaresultsASCIIfile.
(cid:82)∞ x(h)P(h)dh The molecfit package then outputs a transmission spec-
X(h0)= h(cid:82)0∞ TP((hh))dh (8) tsriuomnsapnedctrtuhme doivveirsitohnefoufllthspeeicntrpaultrasnpgecetroufmthebyinpthuetstpraencstrmumis-.
h0 T(h) Apart from results tables in ASCII and FITS format, the cor-
ifthemixingratio(ormolefraction) x,airpressureP,andtem- rected spectrum is written into a file in the same format as the
peratureT aregivendependingonaltitudehbyanatmospheric inputspectrum(seetheUMformoredetails)
profile.Volume-mixingratiosareusuallygiveninpartpermil- In addition, molecfit can provide a new wavelength cali-
lioninvolume(ppmv). brationbasedonthetelluriclinesasdescribedinSect.3.8.This
For H O, it is also common to provide the column height two-stepapproachgreatlyimprovesthespeedoftheoverallpro-
2
ofPWV inmm (alsoreferred toasamount ofPWV). Itcan be cess.
calculatedby
m (cid:90) ∞ x (h)P(h) 3.13. Expertmode
PWV= mol,H2O H2O dh, (9)
ρH2OR h0 T(h) Inmostcasesmolecfitisabletodeterminethedifferentuser-
selected parameters automatically with a minimum of configu-
where the mole mass of water m = 0.0182kg, the den-
mol,H2O ration. However, in a few cases, mostly for CRIRES, a much
sity of liquid water ρ ≈ 103kgm−3, and the gas constant
R = 8.31446Jmol−1KH−2O1. The required units of x, P, T, and h greater flexibility is needed. An expert mode is therefore also
provided. It allows the user to fit the spectra on each inclusion
are ppmv, Pa, K, and km, respectively. Molecfit outputs the range individually and provides him/her the access to the coef-
columnheightforothergasesusingasimilarformula.
ficients of the polynomials for the continuum and wavelength
Since the GDAS profiles and the ESO MeteoMonitor data
correction.Moreover,theparameterfilewiththebest-fitvalues
provideanH Ovolume-mixingratioasrelativehumidityRHin
2 forallparametersissavedandcanbeusedforanotheriteration
percent, it has to be converted to x. For this purpose, we have
ofmolecfit.
implementedtheapproximationsforvapourpressuresoficeand
supercooled water as function of temperature as described by
Murphy&Koop(2005).Theminimumofboththencorresponds 3.14. Fittingskyemissionspectra
to the saturated vapour pressure P , which is used to estimate
sat The tool is able to fit sky emission spectra for molecules and
thealtitude-dependentxinppmvby
physical processes handled by the radiative transfer code, with
P (T(h)) RH(h) theexclusionoflinesproducedbychemiluminescence(suchas
x(h)=106 sat . (10)
OH,see,e.g.,Nolletal.(2012)).Inotherwordsmolecfitcan
P(h) 100
Articlenumber,page8of18
A.Smetteetal.:Molecfit:Ageneraltoolfortelluricabsorptioncorrection
inprincipledeterminethelinespreadfunctionandcolumndensi- such astrategy isshown belowin Sect.6.8 andis illustrated in
tiesoftherelevantmoleculesandthereforederivethetransmis- particularinFig.12.Anotherpossiblestrategyistoensurethat
sion spectrum of the atmosphere based the sky emission spec- thespectralsetupincludesspectralregionscontainingrepresen-
trumifthespectralrangeislocatedinthethermalinfrared. tativetelluriclinesbutfreeofintrinsiclines.
4. Impactonobservingstrategy 4.4. Emissionlinespectra
Although powerful, molecfit is not necessarily suited to tel- Thetellurictransmissionspectrumcannotbefittedonthespec-
luric corrections in all cases. In this section, we briefly discuss trumofanobjectshowinglittleornocontinuum,suchasafaint
afewsituationstoconsiderwhiledecidingonthedetailsofthe cometaryspectrumoranemissionlineobject.Asuggestedstrat-
observingstrategyfortheabsorptiontelluriccorrection. egyinthiscaseiseithertoretrieveaspectrumofatelluricstar
fromthearchiveorobserveatelluricstarinthespectralset-upof
interest.Itsspectrumcanbemodelledbymolecfit.Themodel
4.1. Maincases can then be modified easily by changing the airmass such that
it corresponds to the airmass of the science target. If the water
Twomaincasesfortheuseof molecfitcanbeconsidered,de-
vapouratthesiteismonitored,themodelcanalsobemodified
pendingonwhetherthewavelengthcalibrationofthespectrum
accordingly. The number of times telluric stars need to be ob-
isreliableandaccurate(root-mean-squaredresiduals≈0.1pixel
servedisthereforereducedsignificantly.
orbetter)ornot.
In the case of accurate wavelength calibration, the main re-
quirement is that the spectrum to be corrected has at least one
5. Limitations
spectralregionthatincludesmedium-strengthmolecularlinesof
themostvariablerelevantspeciesandthattheselinesareisolated In this section, we briefly discuss the various factors that limit
enoughsothattheircolumndensityandthelinespreadfunction theaccuracyofthemolecfitcorrectionoftelluricspectra.Such
canbedetermined. factorscanbeexternaltomolecfit,andarerelatedtotheaccu-
If the wavelength calibration is not reliable but needs to be racywithwhichinstrumentalparametersaretakenintoaccount
corrected by molecfit based on the telluric lines themselves, orareinternaltothemethoditself.
a – possibly delicate – selection of inclusion and exclusion re-
gions must be carried out, such that only telluric lines appear
5.1. External
intheinclusionregionsandthatanyintrinsicfeatureoftheob-
ject is masked out in the exclusion regions. In case the contin-
Molecular parameters’ accuracy and completeness: A first
uumneedstoberefinedbymolecfit,thefittingregionsneedto
limitation arises owing to the completeness of the incorporated
presentenoughcoveragetobeabletodetermineeitheranoverall
line database and on the accuracy of the contained molecular
continuumorlocalcontinua.
parameters. Currently, the AER line database delivered with
the LBLRTM code is included by default with molecfit. It
is usually more frequently updated than the underlying main
4.2. Simplecontinuumshapeandsmallnumberofintrinsic
HITRAN11 and is therefore recommended. However, although
features
molecfit a priori allows one to use the original HITRAN
Theeasiestspectrathatcanbemodelledbythetoolaretheones database, this functionality is subject to the stability of the
for objects with a well-defined and simple continuum and with databasestructure.
limitedconfusionbetweentelluricandintrinsicfeatures.Insuch
conditions,thetoolcanproperlymodelthecontinuumspectrum
of the object, and the parameters of the telluric features can be Molecules in atmospheric profile: Molecfit only offers the
possibility to fit the overall volume mixing ratio along the line
wellconstrained.
ofsightintheatmosphere:theoverallshapeoftheprofileisnot
adjusted. If the actual profile differs from the modelled one in
4.3. Largenumberofintrinsicfeatures someatmosphericlayers,systematicerrorscanoccur.Inpartic-
ular such a situation can be caused by the specific, temporary
If the number of intrinsic features is large in the spectral do-
volume-mixing ratios of a molecule and is relatively frequent
mainofinterest,severaloptionsarepossible.Thechoiceofthe
for water vapour. Figure 2 illustrates the possible impact in an
best option depends on the line crowding: a first constraint is
extremecase.
thenumberofpointsthatcanbeusedtodeterminethespectral
In particular, the accuracy of the transmission spectrum de-
continuum;asecondconstraintisthenumberofspectralranges
rivedfromskyemissionasdescribedinSect.3.14stronglyde-
where telluric lines can be used to determine the column den-
pendsontheprecisionwithwhichtheretrievedatmosphericpro-
sity of the molecules contributing in this spectral domain. An
fileactuallyrepresentsthetrueone.
additional constraint arises if the wavelength calibration is not
reliable.
Thereforeapossiblestrategyisfirsttoobserveatelluricstar Atmospheric profile stability: Any correction of telluric ab-
with the same setup – or retrieve corresponding data from the sorptionrequiresarelativelystableatmosphere.Thearrivalofan
archive–anduse molecfitonitsreducedspectrumtodeter- atmospheric front is a clear case where the atmospheric profile
mine which molecules are relevant, the wavelength calibration changessuddenly.EveniftheGDASprofileswouldshowsucha
orcontinuumdetermination.Then,molecfitcanbeappliedto front,itsprecisetimingandcharacteristicsareusuallyuncertain
the science target using the derived wavelength calibration (if andsurelydependonthepointingdirection.Fortunately,fewop-
relevant),changingtheairmass,orfittingthecontinuumand/or
thecolumndensitiesofthemoleculesinvolved.Anexampleof 11 http://www.cfa.harvard.edu/HITRAN/
Articlenumber,page9of18
A&Aproofs:manuscriptno.molecfit
ticalorinfraredastronomicalobservationsareexecutedinsuch 5.3. Internal
conditions.
Good internal guess values: By essence of the Levenberg-
Marquardtfittingapproach,molecfitneedsandissensitiveto
Radiativetransfercodeaccuracy: Thereaderisreferredtothe theinitialguesssolution.Inparticular,molecfitisdesignedto
publicationlistavailableontheLBLRTMwebpage12 forrefer- refinethewavelengthcalibrationandcontinuumdetermination,
ences regarding the validation of the radiative transfer code. In but it is not designed to measure these values from an uncali-
particular,itisworthrepeatingherethat“thealgorithmicaccu- bratedspectrum.Adequateinitialestimatesimprovetherobust-
racy of LBLRTM is approximately 0.5% and the errors associ- nessofthefitandthespeedoftheconvergenceprocess.Knowl-
ated with the computational procedures are of the order of five edgeofthelinespreadfunctionandspectralresolvingpowerare
timeslessthanthoseassociatedwiththelineparameterssothat alsoimportant,althoughthelattercanbeadjustedbymolecfit.
the limiting error is that attributable to the line parameters and
thelineshape”.
Finite pixel size: Molecfit results are best when the spectral
resolutionisatleastNyquistsampledbytheinstrument.Inany
5.2. Instrumental case, molecfit takes the finite pixel size of the spectrum into
account: molecfit estimates the highest pixel resolution in a
Grating scattering: Internal diffusion strongly limits the spectrum by evaluating the wavelength grid. An oversampling
achievableaccuracyofthemodelling.Villanuevaetal.(2009,in- factor of 5 is then applied on the input to the radiative transfer
ternalcommunication)foundthattheCRIRESlinespreadfunc- code.
tion is best reproduced by a Voigt profile that needs to be cal-
culated over ≈ 1000 pixels to account for the grating scattered
light within the instrument. As a consequence, it often appears 6. Examples
thatthecoreofheavilysaturatedabsorptionlinesdoesnotreach
In the following, we illustrate the versatility of molecfit
zero intensity. The amount of flux in the core actually depends
through its application to data sets obtained with various ESO
onthespectralenergydistributionofthelightreachingthegrat-
ing and cannot be modelled accurately by molecfit, because VLT instruments covering a range of spectral resolving powers
molecfitonlyappliestheconvolutionofthelinespreadfunc- andwavelengthdomains.
tion to the transmission spectrum and not to the science target
one. In addition, later investigation (Uttenthaler 2010, private
6.1. UVES
communication) indicates that the spatial profile of the grating
scatteredlightisdifferentfromtheoneofthedirectlight. The prototype version of molecfit was manually adjusted
The observed spectrum therefore combines different effects to determine the location and strength of telluric absorption
that cannot be easily modelled by molecfit. Therefore they lines in VLT UV–Visual Échelle Spectrograph (UVES, Dekker
limittheaccuracyofthecorrectionsuchthatresidualscanreach et al. 2000) Rapid Response Mode or Target-of-Opportunity
anestimated1%ofthecontinuumdependingonthesignificance programmes of γ-ray bursts (UVES is a high-resolution cross-
ofthegratingscattering. dispersed spectrograph in the visible): indeed, some H2O lines
mimic lines associated with high-redshift damped Ly-α (DLA)
lines. For example, in the spectrum of GRB050730 (Ledoux
Instrumental background: When modelling sky emission et al. 2009) obtained under Programme ID 075.A-0603 (P.I.:
spectra,molecfitusesabackgrounddefinedbyagreybodyat Fiore),H OlinesappearcoincidentinwavelengthwiththeSiii
2
the temperature of the main mirror of the telescope. It assumes 2P λ1816line,andaffecttheFeii6D λ1267andFeii4F
3/2 7/2 7/2
thatthethermalbackgroundfromtheinstrumentisnegligible. λ1659lines,allassociatedwiththez=3.69857DLA.Similarly,
as shown in Fig. 4, using molecfit allowed the location and
strengthof theH Olines affectingthe Feii λ2382, λ2374,Feii
Instrumental response curve shape: Molecfit fits a low- 2
4D ,andλ2389linestobedetermined(amongstothers)associ-
orderpolynomialtotheobservedcontinuumoftheinput(object 7/2
atedwiththez = 2.42743DLAinthespectrumofGRB080310
spectrum)foreachinclusionregion.However,itdoesnotmake
a difference between the intrinsic continuum shape and the ef- (DeCiaetal.2012)obtainedunderprogrammeID080.D-0526
(PI.Vreeswijk).
fect of the response curve of the instrument. Good results are
obtainediftheobservedcontinuumofeachinclusionregioncan
be modelled well by such polynomials. Alternatively, the user 6.2. FLAMES
mayprovideanormalisedspectrumasinput.
Figure 5 presents a Fibre Large Array Multi Element Spectro-
graph(FLAMES)/GiraffespectrumoftheO4VstarNGC3603-
Linespreadfunction: Thelinespreadfunctionshouldbehave 117observedwiththeLR8settingandtheARGUSintegralfield
welloverthewavelengthrangeofinterest,eitherbeingconstant unit(datacourtesyofM.Gieles;Prog.ID:079.D-0374(A)),pro-
or having a width that is linearly dependent on wavelength, as vidingaspectralresolvingpowerofR ≈ 10400.Thepresented
expectedincross-dispersedspectrographs. spectrumcorrespondstothebrightestspaxelontheobject.
When the transmission spectrum is derived from the sky The presence of water-vapour telluric absorption lines im-
emissionspectrum(Sect.3.14),oneshouldnotethattheparam- pedesanyprecisedeterminationofthewidthandcentroidofthe
eters of the line spread function are likely to be different from Paschen hydrogen lines at 901 and 923 nm. However, there is
theoneofthesciencetargetgiventhattheimageofthestardoes usuallynoobservationoftelluricstarsforFLAMES.Molecfit
notuniformlyilluminatetheentranceslit. can be used to determine the PWV and to correct the science
spectrumbythetransmissionspectrum.Asaresult,aprecisede-
12 Inparticular,http://rtweb.aer.com/lblrtm_ref_pub.html terminationofthewidthandpositionofthelinescanbeobtained
Articlenumber,page10of18