KSPECTRUM version 1.2.0 User Manual
Contents
1. X Xo X95 X3 Figure 2 Purpose of tabulating function f x for any x 21 29 the maximum relative difference between f and f is lower than the imposed value ev In practice for the computation of the spectral grid starting from a wavenumber value v the algorithm will have to determine the next grid point vi see Fig 1 For every line l whose center wavenumber is within Vmin Vmax the code will have to identify the triplet 11 21 73 so that DES a 11 21 We know that the position of the next grid point could safely be taken at x3 if line l was alone Taking the minimum value Vier min Yigis YL1 31 Ve over all lines present in the interval ensures that the relative difference between ka and k never exceeds gt over Vi Viz _42 The principle is exactly the same for the Voigt function fy v a 7 4 ae 7 lt DE dt with z yIn 2 y yIn 2 and yp the Doppler linewidth This function depends on y the ratio of Lorentz and Doppler linewidths This function will therefore have to be tabulated as a function of z but also as a function of y I really don t see the point in giving any more details If anyone ever reach this point without falling asleep and reads this please send me an email at vincent eymet meso star com A special mention should be made tough to the issue encountered for meshing the spectral grid between close strong lines In the case v is lo2. The user can then choose between the CDSD 1000 the CDSD 1000 updated the CDSD HITEMP or the CDSD 4000 databases for CO2 and between HITEMP old version or HITEMP 2010 for H20 for a temperature greater than T_hitemp whose value was set in the data in file Section Composition the user should specify whether the code will use only the main isotope or all isotopes when the concentration of isotopes are not specified for a given molecule Section Line profiles first the user can select to use either the Lorentz or the Voigt line profile the user has then to chose whether or not the sub lorentzian nature of lines has to be taken into consideration See section 2 3 for more information about how the code will manage sub lorentzian profiles In particular function x is given for pure CO only in specific spectral ranges and for a limited range of temperature The user can chose to use literature data outside its range of validity over the whole IR spectrum and for other molecules than COo use with great care Finally the asymmetric nature of function y can be taken into account but results will not be very different from using a symmetric function Section Collision Induced absorption the use can chose whether or not collision induced absorption CIA should be simulated for COg See section 2 3 The user can subsequently choose CIA computation from the work of Gruszka 16 or Baranov 19 or both
3. In s Supplemental 35_hit08 par In s Supplemental 42_hit08 par 21 4 Using the code 4 1 User input files There are two files in the main KSPECTRUM folder data in and op tions in The options in file is where the user can chose between several options for each algorithm The data in file is where the user can specify the numerical values of a number of variables the code will use Below is a description of these two files 4 1 1 The data in file e Section Production of k spectrum the user must specify the values of and ez see section 2 1 Default values are 1 e Section Narrowband discretization the user can specify the mini mum and maximum values of the wavenumber range the spectrum will be computed for and the maximum error percentage over the Planck intensity for discretization of the wavenumber range into narrowband intervals These data should be specified only if the user chose to com pute the narrowband discretization instead of using pre defined nar rowband intervals via an option in the options in file Refer to the description of the corresponding option for more details Finally the user should specify the value of the constant spectral step to use when required in options in e Parallel computing the user should specify the number of chunks the code will use for the computation of the k spectrum Once the spectral grid is c
4. 4 107 1 1 107 2745 2790 2 0 107 4 0 1073 2790 2925 7 9 1073 2 5 1072 2925 3040 2 2 10 1 1 1 107 3040 3225 2 2 107 9 9 1073 3225 3450 2 2 0 18 3450 3760 604 6 592 2 3760 3875 28 2 19 4 3875 4030 0 17 9 4 1072 4030 4135 4 1 107 1 9 107 4135 4350 3 2 1073 1 1 1073 4350 4550 6 8 1078 3 3 1074 4550 4950 4 8 3 9 4950 5200 10 2 9 4 5200 5700 2 7 107 1 6 107 5700 5825 3 4 1078 1 6 1073 37 References 1 L S Rothman D Jacquemart A Barbe D C Benner M Birk L R Brown M R Carleer C Jr Chackerian K Chance L H Coudert V Dana V M Devi J M Flaud R R Gamache A Goldman J M Hartmann K W Jucks A G Maki J Y Mandin S T Massie J Orphal A Perrin C P Rinsland M A H Smith J Tennyson R N Tolchenov R A Toth V J Auwera P Varanasi and G Wagner The HITRAN 2004 molecular spectroscopic database Journal of Quantita tive Spectroscopy and Radiative Transfer 96 139 204 2005 L S Rothman L E Gordon A Barbe D C Benner P F Bernath M Birk V Boudon L R Brown A Campargue J P Champion K Chance L H Coudert V Dana V M Devi S Fally J M Flaud R R Gamache A Goldman D Jacquemart L Kleiner N Lacome W J Lafferty J Y Mandin S T Massie S N Mikhailenko C E Miller N Moazzen Ahmadi O V Naumenko A V Nikitin J Orphal V I Perevalov A Perrin A Predoi Cross C P Rinsland M Rotger M S
5. calculation_info txt will be needed when producing k distribution data sets e optimizations LBL files shows which LBL data files have been read e status txt is continuously updated during computation Some lines are appended to this file each time a backup is performed i e when the computation of the spectra is over for a given narrowband spectral 5 3 28 interval It shows the time the backup took place at for what atmo spheric level narrowband interval and how many values of k have been computed for this interval When running a multi pass compu tation it can also display the predicted time date when computation will be over for the current atmospheric level What s next You can have multiple reasons to use KSPECTRUM Produce your own high resolution spectra and use them as you wish Produce a spectra database that can later be used for interpolation purposes we have a code for that Perform high resolution radiative transfer often refered as line by line computations We can provide example codes Produce statistical spectral model parameters for instance correlated k datasets We have a code for that And then you might want to perform radiative transfer computations from this correlated k data We can also provide example codes In any case do not hesitate to contact M so star contact meso star com 29 6 The fun stuff 6 1 Legal mentions KSPECTRUM is distributed under th
6. defined narrowband intervals If the narrowband discretization has to be per formed by the code it will use the three corresponding values in the data in file section Narrowband discretization the specified wavenum ber range will then be discretized into a number of narrowband inter vals and the Planck intensity relative variation between the limits of each narrowband interval will not exceed the specified maximum er 23 ror percentage over Plank intensity If the user decides to use a pre defined narrowband discretization a file called narrowbands in must be found in the data directory This file should contain the number of narrowband intervals and the wavenumber limits of each interval See the provided data narrowbands in example file Section Choice of spectroscopic database A first option makes pos sible to use custom LBL databases parts from existing databases for instance The corresponding LBL files have to be written using the HI TRAN format These files must reside within the data custom_LBL folder and must be named according to the molecular species the tran sitions they contain belong to for instance custom_H2O txt or cus tom_CO2 txt When custom LBL databases are not used the user has to decide whether the default HITRAN database to use is the 2004 or the 2008 version the corresponding link must be established within the data folder see section 3 5
7. in the newly created directory e Go into the data directory compile program make_data exe and 9 6 2 run in in order to generate test files composition in narrowbands in and molparam in see section 4 2 e Remind you to install LBL databases into the data directory see section 3 5 Then you can go into the newly created directory named kspectrumX X X with X X X the version number to continue the installation 2 Description of the code 2 1 Production of high resolution spectrum full con trol of the uncertainty KSPECTRUM has been designed to produce high resolution spectrum of any gas mixture in any thermodynamical conditions from line by line databases In practice it is far from being the case see sections 2 2 and 2 3 mostly be cause we have no idea of sub lorentzian profiles and collision induced absorp tion CIA in the general case This code will be improved gradually when our knowledge of spectroscopy evolves Anyway the purpose of the present code is clearly NOT to produce spectrum that will be used in engineering applications i e spatial missions Rather it is intended at producing re liable spectrum that may be used for subsequent radiative transfer analysis production of radiative transfer parameterizations into planetary GCMs By high resolution spectrum we mean that the spectral resolution is high enough so that individual lines are well resolved
8. in which case 24 both sources of opacity are added The user can also chose to compute CIA coefficients outside the wavenumber validity range 0 250 cm7 and the temperature validity range 200 800 K Section Code behavior the user can chose to display or not com putation times on screen Finally it should be specified whether the code has to resume an interrupted computation or start again from the beginning Section Sensitivities the user can chose to compute sensitivities of results cross section and absorption coefficient to total pressure tem perature and species concentrations When computation of sensitivities are enabled results are recorded in files results dk_dXXX Section Levels and narrowbands limits the user must chose whether the code has to compute a spectrum for every atmospheric level nar rowband interval or only for a limited number of levels intervals If a limitation is imposed the code will use the corresponding values found in the data in file section Levels and narrowbands limits Section Spectral discretization algorithm the user should choose be tween the reference spectral discretization algorithm a degraded spec tral resolution using a constant spectral step or using a spectral dis cretization based on line centers Degraded resolution will compute the spectral grid faster than the reference algorithm and will also make the computation of the
9. is not acceptable since KSPECTRUM is sup posed to be used for any gas mixture in any spectral region and in a wide range of thermodynamic conditions However such a code would require a good knowledge of all the missing data sub lorentzian profiles CIA tran sitions and broadening by the rest of the gas mixture for every molecular species broadened by every other molecular species for any spectral region for every temperature and pressure level The best I can propose is to im prove the code for very specific applications when the corresponding spectro 31 scopic data is known Please let me know your needs vincent eymet meso star com A not fully satisfying solution can be found in the case you know well the CIA continuum you want to take into account i e you know how to compute CIA opacities for your configuration but it has not been included into KSPECTRUM in this case you should first compute your high reso lution spectra without any CIA calculations Once high resolution spectra without CIA have been computed you can compute and add CIA opacities by yourself at each v point 32 A Appendix Validation KSPECTRUM results have been checked in various situations against other results A 1 Low pressures temperatures Results from KSPECTRUM have been compared to line by line spectrum produced by P Dubuisson Laboratoire d Optique Atmosph rique Lille 20 21 22 23 24 for Earth s atmopshere Various co
10. measurements from Tonkov at al 11 in the 2 3 um region at room temperature Central wavenumbers and intensities are given for 8 transitions but the authors clearly state that the data they provide should be considered with caution because of the large uncertainties of their measure ments computations Other measurements of CO CIA have been reported in the literature 12 13 14 15 Data lines central wavenumber and inten sity are either not provided or given with poor accuracies or provided in a very limited range of validity in very specific spectral regions for only one value of pressure and temperature etc Practically it is not possible to use literature results to take into account collision induced or pressure induced absorption lines properly The only results that could practically be used for computing CO 2 CIA is the work from Gruszka and Borysow 16 In addition to the paper the authors provide a fortran computer code 17 that will compute CIA for CO in the 0 250cm wavenumber range in the 200 800K temperature range this was motivated by studies of Venus atmosphere since CIA is a dominant source of opacity in the far infrared between 0 and 550cm7 on Venus atmosphere 18 where temperature reach 735K at ground level This code has been modified for computing CIA coefficient for only one value of v and integrated into KSPECTRUM CIA of CO will therefore accurately be taken into account within the
11. of the second million of lines to the final result And so on until all lines have been taken into account An attentive reader could tell me this approach is incorrect if applied to the spectral discretization Indeed in order to compute a correct spectral grid for any Vmin Vmax narrowband spectral interval the algorithm needs to take into account ALL the lines whose central wavenumber is within Vmin Vmaz When reading chunks of LBL database it is obvious that line parameters are missing for many Vmin Vmax intervals This is the reason why the spectral discretization algorithm does not use the multi pass approach it automati cally looks into every LBL database file to identify relevant line parameters for each Vmin Vmax interval in practice a complete parsing of LBL data files is necessary in order to accelerate the process of line identification This is also why LBL data files must be readable by any KSPECTRUM process see section 3 5 Finally we should mention that the total computation time for a multi pass computation is not higher than for a 1 pass computation since all lines have to be taken into account anyway The spectral discretization is per formed during the first pass and is known for subsequent passes which does not increase the total computation time The main advantage of this multi pass approach is the possibility to run the code on low memory configura tions 2 7 Estimation of remaining computation ti
12. spectrum faster since less points will have to be computed However the accuracy criterion 2 will not be satisfied by any other algorithm than the reference scheme This degraded res olution algorithm takes input values see data in When using the reference spectral discretization algorithm the user should also pro vide the accuracy level that will use the special spectral grid algorithm for wavenumbers located between close strong lines A value of 0 dis ables this special algorithm wavenumbers v will be computed using the standard meshing algorithm based on Lorentz and Voigt functions tabulation described in 2 but remember that in this case the spectral grid may be too coarse between strong lines A value between 1 and 4 will enable the special meshing algorithm The highest the value the more accurate this algorithm will be there will be more values of vj therefore making the k spectra computation longer A value of 3 may be enough in most cases Section line truncation and selection the user has the ability to trun cate line profiles at a fixed distance from line centers or for a multiple of 25 the line width see data in When this option is used the accuracy criterion is no longer satisfied Line truncation will make kspectrum computations much faster When a truncation is required the user has also the ability to set the contribution of constant lines to zero so that absorption is null between
13. the data composition in file and note that molecule labels must be provided between brackets example H gt O so that the code can identify molecular species 4 2 2 Isotopologue abundances For each molecular species that is provided in the data composition in file i e molecules whose label is unknown from the HITRAN nomenclature information has to be provided about these molecules in the data mol param in file In particular the abundance of each known isotopologue should be provided for each P T level See the provided example file pro duced by the program that can be found in the data make_composition for 26 source file One remark about isotopologues when no isotopic abundance has been explicitly provided via the data molparam in file for a given molecular species the algorithm can chose to take into account only the main isotope of this species or use all isotopes whose line parameters are provided in LBL databases The user can chose between both behaviors in the options in file see section 4 1 2 4 2 3 Narrowband intervals have to be provided in the data narrowbands in file provided that the user chose to use a pre defined narrowband discretization see section 4 1 This file must contain the number of narrowband intervals and for each interval its lower and upper limits in terms of wavenumber in cm 4 3 Running KSPECTRUM using MPICH or OPEN MPI Once ev
14. very distant lines A last option enables weak lines rejection in order to speed up cal culations the computation can be performed using only the stronger lines those who have the highest intensities This option enables the convergent k algorithm that sorts lines according to their line intensity computed at the current mixture s temperature and adds the contribution of intense lines first until the convergence criterion is reached This criterion can be adjusted using the corresponding value in the data in file 4 2 Other input files 4 2 1 Atmospheric composition is provided via the data composition in This file must contain a label for the atmospheric composition could be used in future version for planet specific physics such as sub lorentzian profiles or CIA the number of atmo spheric levels the number of molecular species the label of each molecular species and for each atmospheric level altitude in km unused at present date pressure in atm temperature in Kelvin and the molar fraction or partial pressure of each molecular species See the example file Molecule labels have to be identical to labels used in the HITRAN nomen clature Labels of molecules defined in the HITRAN database can be found in the data molparam txt file They are obvious for most species H20 is used for water vapor CO2 for carbon dioxide etc Please use capi tal letters when providing molecule labels in
15. will be able to compile KSPECTRUM with this value you need at least 1GB RAM You can enable a lower value if you do not have enough memory The code will then perform a multi pass computation see section 2 6 but total computation time will not be increased 3 3 Compilation Once you checked compilation options and array size definitions you can use the make all command in order to compile the executable file If compilation fails use the compiler error message to determine what went wrong The most probable error causes are a bad definition of archi tecture compilation option or an inappropriate value in code optimization options 16 If you ever need to modify the source files in directory source you can quickly recompile the code using make all again This will only recompile the modified source files and link objects files in order to produce the new executable file If you have to modify the value of any variable defined in includes files di rectory includes you will have to recompile the whole code from scratch Use the make clean all command to erase all objects files and then recom pile them properly Odd errors may happen if you change an include file and then recompile using only the make all command old value of the modified variable will remain in the unchanged object files 3 4 Installation of MPICH If you do not already have MPICH installed on the machine group of machines you want to ru
16. KSPECTRUM version 1 2 0 User Manual February 4 2015 V Eymet M so star www meso star com LAboratoire PLasmas et Conversion de l Energie Universit de Toulouse Legal mentions KSPECTRUM is a free software released under the terms of the CeCILL license This license is compatible with the GNU General Public License GPL See the COPYING file within the main directory M so Star www meso star com provides and distributes the sources of the present code By using this software you agree to make a reference to KSPECTRUM in any publication using results that have been obtained with it Support for the present code is no longer free If you need to adapt the code to your needs for any computational request or if you need to get training in order to use the code please contact M so Star contact meso star com and we will provide an estimation of our fees Contents 1 Getting started 3 2 Description of the code 3 2 1 Production of high resolution spectrum full control of the un GEE ALT Ga 96 tos loa Se tes ks yok eae e GPA eee Gs aE ee ef es ae 2 3 2 2 Range of applications o ey dah ne ce ede eves bey we a pee 7 2 3 Physical hypothesis tar be eile ee Wits olen ei eae oh he be A B 8 2 3 1 Sub lorentzian profiles and CIA 2 3 2 Broadening by the rest of the gas 2 3 3 High temperature transitions 2 4 A better respect of spectroscopy 4 2 4 1 no line s
17. OA 011 0 00012 kspectrum 01 4 ssid 4 kspectrum O11 see 0 0001 0 0001 F i i J 8e 05 a E 1e 05 i i 4 6e 05 x i Iy 1e 06 2e 05 0 i i 1e 07 i i 4214 4215 4216 4217 4218 4219 4220 4214 4215 4216 4217 4218 4219 4220 v cm v cm Figure 5 Absorption spectra for a H20 CO2 O3 and C H4 mixture at pres sure of 0 259 atm m o 1 74 1074 zoo 3 30 10 ro 1 54 1077 xc 1 56 10 and a temperature of 232 1K in the 4214 4220 cm7 spectral range 36 Table 1 Average absorption coefficients k computed from KSPEC TRUM high resolution output spectra and from k distribution data used for radiative transfer simulations 25 for some narrowband spectral inter vals at ground level 90 1 atm 735K Values of ka extracted from the k distribution data set are not a reference but are rather used as a comparison point Ving Vsup ka k distributions ka KSPECTRUM 520 570 60 5 48 8 570 615 574 1 529 8 615 645 2 36 103 2 21 103 645 680 7 56 103 8 02 103 680 720 3 71 103 3 74 10 720 760 899 3 913 1 760 785 205 3 179 9 785 815 58 4 42 2 815 845 4 0 8 5 845 900 8 9 5 1 900 935 5 7 13 9 2100 2155 4 5 5 9 2155 2280 3 22 103 2 56 103 2280 2390 5 28 104 5 42 104 2390 2450 1 48 103 1 06 103 2450 2540 5 9 10 6 2540 2670 2 8 107 3 1 107 2670 2745 1
18. This point is dis cussed below The main purpose of producing high resolution spectrum is the possibility to compute k distribution data sets The main idea that lead to the development of KSPECTRUM is to pro duce a code that can compute high resolution data for virtually any condi tions with a full control of the accuracy e Any value of k v has to be computed with a specified relative error E1 e The maximum relative error made when considering a linear profile k v between two computed values k v and ka Vi may also be specified gt see Fig 1 Ideally when computing the absorption coefficient k v at a given wavenum ber v the contribution of every line should be summed up kav 5 karv 1 where ka v is the contribution of the N lines index to ka v In practice it is impossible to use such a method to compute k v at high spectral resolution because the order of magnitude of N is typically several hundreds of thousands when not millions The first constraint of accuracy will be used to set up a faster algorithm we start by discretizing the infrared spectrum in a given number M of nar rowband intervals whose spectral limits are known Let us take the example of any spectral interval Vmin Vmax Each one of the N lines will be exam ined for this interval For a great number of lines Vmin Vmax Will be located in the far wings and therefore the values of ka Umin and ka Vmax are not e
19. a huge ugly bug for every 1000 lines of code My code counts 20 klines of code So yes there may be a bug Send me an email with the description of what you want to do attach input files when possible Error Nkm x has been reached This is one of the most likely errors You are probably trying to compute a spectra for a very low value of pressure right There are a lot of very thin and isolated lines the code takes ages to compute the spectral discretization and then crashes with this error This is because the number of wavelength the spectra should be computed for is too high There are a number of ways to get around this issue e First try to edit the includes max inc file and to increase the value of variable Nkmx in order to allow more memory to spectral arrays 30 You will have to recompile the code completely to take the modification into account use make clean and then make all or directly make clean all Be very careful when increasing the value of Nkmx if you do not increase it enough the same error will happen again not enough memory for spectral discretization If you increase it too much you will not be able to compile the code not enough available memory on the system The default value for Nkmx is 10 you should first try to set it at a value of 5 10 Then increase it if the same error occurs again until you reach the limit at which compilation and or executio
20. cated between two strong lines the value of v computed by the above described algorithm may be too far from 1 to satisfy accuracy criterion over 2 because of the strong overlap between the two lines In this case a special treatment should be applied for a correct discretization of wavenumbers 2 2 Range of applications KSPECTRUM is currently using the following LBL databases HITRAN 2004 and 2008 1 2 HITEMP 2004 3 H20 CO2 CO and OH only HITEMP 2010 4 H20 CO2 CO NO and OH only and CDSD 5 CO only The code uses the HITRAN database for temperatures lower than a user defined temperature T_hitemp defined in the data in file see sec tion refpara userfiles For temperatures greater than this value it lets the user chose between HITEMP or CDSD for CO lines and will automatically use HITEMP for H0 CO and OH transitions The temperature upper limit of validity of the present code is therefore the same as the HITEMP database about 1000K except when computing the spectrum of a pure CO2 atmosphere the CDSD database is supposed to be accurate up to 3000K No physical limit is imposed for pressures However the upper limit in terms of pressure is probably what is found at ground level on Venus ground pressure is 92 bars and ground temperature is 735K At such levels of pressure and temperature the atmosphere is in super critical thermody namic state and can no longer be considered as a mixture
21. ck atmosphere of Venus e Collision induced absorption CIA that are mainly due to the fact that in high pressure and or high temperature conditions collisions between molecules temporarily create new molecular species with their own energetic transitions Sub lorentzian profiles have been studied since 1969 in 6 the Lorentz profile f v vo is corrected by a function y v vo that accounts for the sub lorentzian nature of the lineshape Function y v vo is given for pure CO and for mixtures of CO and other species N3 He Ar O2 H3 in very specific spectral ranges for various values of the temperature In 7 and 8 function xy is given respectively for pure CO and for CO2 broadened by O2 and No in the 4 3 um region for different temperatures It was then shown in 9 that function y is asymmetric with respect to vo for CO in the 4 3um region at 296K Finally Perrin and Hartmann 10 and Tonkov et al 11 provide x func tions for pure CO respectively in the 4 3 um region for T 190 800 K and in the 2 3m region at 296K These results are used in KSPECTRUM in order to compute x profiles Various options see section 4 let the user chose whether x functions from the literature must be used in their strict range of validity or if they have to be used in different spectral ranges for other molecules than CO Collision induced transitions are even more difficult to take into account there are some
22. e terms of the CeCILL license This license is compatible with the GNU General Public License GPL See the COPYING file within the main directory M so Star www meso star com provides and distributes the sources of the present code By using this software you agree to make a reference to KSPECTRUM in any publication using results that have been obtained with it Support for the present code is no longer free If you need to adapt the code to your needs for any computational request or if you need to get training in order to use the code please contact M so Star contact meso star com and we will provide an estimation of our fees 6 2 Questions and Answers I am in a big hurry I don t want to read all that boring stuff I just want my results immediately KSPECTRUM is probably not what you want Consider using online spectra production systems such as http hitran iao ru gasmixture spectr or http spectra iao ru I can t manage to compile the source code I don t understand compilation error messages There should be not problems using the gfortran compiler and mpich2 or openmpi Send me an email with a description of the tools you are using on what hardware system and what error messages you get vincent eymet meso star com Compilation was successful but the code crashes I know that when a code has been extensively used for years and all bugs have been patiently tracked identified and corrected there still remains
23. election i aye ae Ga ee Oe Banta 2 4 2 no truncation of lines and continuums 2 5 Parallel computing a5 wy There a euty ae PRY iy vie Ths ee ws a Ne es 2 6 The multi pass approach kb ed ee en eid Wd eG Bd 2 7 Estimation of remaining computation time 2 8 Not starting from scratch after a crash 2 9 Automatic tabulation of Lorentz and Voigt functions Installation and prerequisites 3 1 Tweaking the Makefile 4 9 2 08 patos We PaaS as 3 2 Setting up array sizes 2s a als a a ab aa 3 37 Compilation 2 siirat ak pa one he ee he TA a eS 3 4 Installation of MPICH 205 3792 81 08 Y a 3 5 Installation of LBL databases o oaoa aaa aa Using the code 4l User mput Mes tas a peP e a e Geel ard A E a h a aw ak ATI bes sein Mle 32 lt p grie ieg Sirs a d ge ee io 4 1 2 The options in file o aoaaa 42 Other input fless aa oe ee Bee Oe eh a aR 4 2 1 Atmospheric composition 4 2 2 I sotopologue abundances onoo a a de oe 4 2 3 Narrowband intervals ooa a 8k OE BR Od 4 3 Running KSPECTRUM using MPICH or OPENMPI Results Bul Results tiles na kes shag ke Se ee aa ae dee a 5 2 More output files 2 2 4 s0o se Sa Boe Sa eee oe eee 5 3 What s next 2 0 0 0 000000000 eee The fun stuff 6l egal mentions su 2 42 eye ead A a a ar a o tee Ae ase 6 2 Questions and Answers i arose yb Be a eS Appendix Validation A 1 Low pressur
24. elte snl 5 is lower than the ka Vmin ka maz 2 very different If the relative error specified value 1 then a constant value for instance can Figure 1 Schematic representation of the computed k v spectrum and of the maximal error made when considering a linear variation k v between two consecutive computed result points be taken as ka v for every value v Vmin Vmax With a total relative error over k v that will be lower than The second accuracy constraint 2 will be used for optimizing the spec tral discretization schematically narrowband intervals Vmin Ymar will be discretized in such a way that line centers are well described Discretization steps will take greater values in line wings The computation of the spectral grid uses a tabulation of the Lorentz and Voigt functions The method is explained below for a Lorentz line profile For an isolated Lorentz line we have YL SE 2 ATU R with yz the Lorentz line width and ve vo P the corrected wavenum ber at line center This function can be written as 1 1 S 3 f YL 1 z with z Next function f x a has been tabulated in order to determine a series of triplets x1 x2 73 such that for any value xo x1 2 the maximum relative difference between f x and f x f xo Leafe y Xo defined for x a9 3 never exceeds Figure 2 may help Emax lt E
25. erything is installed and the executable file kspectrum exe file has been compiled you can try to run a computation I would recommend that for the first time you run KSPECTRUM over a single atmospheric level for a single narrowband interval using a simple atmospheric composition the provided example composition for instance Use the following command to run the code mpirun np kspectrum exe with the number of processes that have to run Because communication times are small compared to computation times in KSPECTRUM it is a good idea to chose a number of processes equal to the number of physical processors of your cluster plus one One process the master process is dispatching computational loads to every other processes slave processes and gathering results from them It does not require any significant CPU time therefore it is OK to have a number of slave processes equal to the number of processors so that each slave process can use a processor or each processor will have only one slave process running on it In practice if your cluster is composed of n processors you can use mpirun np n 1 kspectrum exe 27 5 Results 5 1 Results files KSPECTRUM will produce results files in the results directory There is one file per atmospheric level Files are named k with the index of the atmospheric level For instance results k001 is the result file produced for the first atmo
26. es temperatures 2000 A 2 High pressures temperatures 00 10 10 10 11 11 11 12 13 13 15 15 15 15 16 17 21 21 21 22 25 25 25 26 26 27 27 27 28 29 29 29 1 Getting started The first thing you should do is to declare your fortran compiler in the F77 environment variable In order to do so you have to export the F77 environment variable in your bashrc cshre or profile file for instance add the following line into your bashrc or profile file export F77 gfortran in case you are not using the gfortran compiler replace gfortran by your local fortran compiler if needed ifort pgf Then you need to type source bashrc in order to take the modifications into account Replace bashrc by whatever you are using The purpose of defining the F77 environment variable is to provide the correct compiler to compilation commands that will be used later by the installation script If this variable is not defined the name of your fortran compiler will be asked every time a compilation will occur You need to use the install_kspectrum bash script that is provided with the archive This script should be placed in the same directory than the provided zipped archive Use the install_kspectrum bash command in order to run it This script will e Untar the archive if the corresponding version of KSPECTRUM is not already installed e Make appropriate links
27. ill never end Fortunately KSPECTRUM can resume a computation if a crash has been detected It will start over from the last backup point which occurs every time the computation of the spectrum has been achieved on a given narrowband interval Multi pass computations will be resumed as well 2 9 Automatic tabulation of Lorentz and Voigt func tions Section 2 1 presents the principle of the spectral discretization algorithm used in KSPECTRUM It needs a tabulation of the Lorentz and Voigt functions These tabulations depend on the value of 2 the user defined accurately that is required over the spectral grid Tabulation files reside within the data directory and are indexed according to the value of gt for instance data tabulation_lorentz0 01 txt is the file where x1 2 3 triplets have been recorded for the Lorentz function and for an accuracy e2 0 01 a 1 uncertainty Values of and g may be changed by the user see section 4 1 If the user specifies a value of 2 for which there is no known file KSPECTRUM will have to perform the tabulation prior to start the spectrum computation Of 14 course the code will record tabulation results in the appropriate file so that the tabulation for this value of e gt will never have to be performed again This tabulation is performed by a fully parallel loop so that it can benefit from the number of processors KSPECTRUM has been launched on 15 3 Insta
28. imeckova M A H Smith K Sung S A Tashkun J Tennyson R A Toth A C Vandaele and V J Auwera The HITRAN 2008 molecular spectroscopic database Journal of Quantitative Spectroscopy and Ra diative Transfer 110 533 572 2009 L S Rothman C Camy Peret J M Flaud R R Gamache A Gold man D Gootvitch R L Hawkins J Schroeder J E A Selby and R B Wattson HITEMP the high temperature molecular spectroscopic database available through http www hitran com 2000 L S Rothman I E Gordon R J Barber H Dothe R R Gamache A Goldman V I Perevalov S A Tashkun and J Tennyson HITEMP the high temperature molecular spectroscopic database Journal of Quantitative Spectroscopy and Radiative Transfer 111 2139 2150 2010 S A Tashkun V I Perevalov J L Teffo A D Byko and N N Lavren tieva CDSD 1000 the high temperature carbon dioxide spectroscopic databand Journal of Quantitative Spectroscopy and Radiative Transfer 82 165 196 2003 D E Burch D A Gryvnak R R Patty and C E Bartky Absorption of infrared radiant energy by CO and H0 iv shapes of collision broadened CO lines Journal of the Optical Society of America 59 267 280 1969 7 12 13 14 15 16 38 R LeDoucen C Cousin C Boulet and A Henry Temperature depen dance of the absorption in the region beyond the 4 3 um band head of CO 1 Pure CO case Applied Optics 24 No 6 897 906 1985 C Co
29. led a line profile truncation possibility in the code see section 4 1 end of options in description This option should be used with great care since it is most probable that very intense and distant lines will not be taken into account when truncating line profiles 2 5 Parallel computing Because of accuracy requirements that are imposed computation times could be prohibitive on a single processor machine KSPECTRUM has been de signed for running on multi processor machines clusters It uses MPICH instructions so that multiple processes can be run at the same time The four time consuming loops of the code have been parallelized As communication times between processes are negligible compared to computation times the total computation time will effectively be divided by the number of proces sors KSPECTRUM is running on provided that all processors machines in the cluster run at the same speed and are free of other time consuming processes Of course this means MPICH has to be installed before using the code See section 3 4 for installing MPICH creating a cluster of machines 2 6 The multi pass approach We are now talking about the limitations of fortran 77 which has been used for coding KSPECTRUM The source code is constantly checked against array overflows unused or uninitialized variables etc classical programming errors and there should be no problem of bad coding However one lim itation of fortran 77 i
30. llation and prerequisites 3 1 Tweaking the Makefile Before compiling you will have to find out what compilation options are right for your compiler and your machine Open the Makefile file and look at variables FOR ARCH and OPTI Variable FOR is used to specify your fortran 77 compiler As KSPECTRUM uses MPICH you will most likely use the mpif77 compilation command that has been installed along with MPICH Variable ARCH is used to specify machine architecture m486 is probably a good choice for a PC running a 32bits linux Use the documen tation of your fortran compiler to find out what architecture option you can use Variable OPTI is used to specify code optimization options The de fault options should be enough Please note that you definitely must use option Wno globals for compiling parallel code You might also want to set variable DEBUG look for its definition in the file You can expect faster execution times if you leave it empty 3 2 Setting up array sizes As explained in the previous section see 2 6 one limitation of fortran 77 code is that you must define array sizes before compilation Arrays sizes used by the present code are defined within the includes max inc file You should at least look at it before compiling and more precisely at the value of variable Nline_mx Its default value is 2000000 2 10 because most machines
31. me When using the multi pass approach the code will be able to guess the remaining time before the computation of a spectra is finished for the current 13 atmospheric level This is only possible for a multi pass computation because in order to have an estimation of the remaining computation time the code needs to know the spectral grid that is computed during the first pass of each atmospheric level and how many points of the grid have already been computed The status txt file will report the estimated time in real life date hour format when the computation for the next atmospheric level is going to start 2 8 Not starting from scratch after a crash Imagine you want to compute the spectrum of a given atmosphere mainly composed of COs using the CDSD database over 80 atmospheric levels for the whole infrared range This should typically require several weeks of CPU time even for a multi processor machine or a small size cluster If you run KSPECTRUM on a dedicated multi processor machine if no one else has access to this machine and if there is no power cut during all the computation there should be no problem Now imagine you run it on a PC cluster with many other users that daily access these machines Someone will eventually reboot one of these machines crashing your whole computation I would say that you can not decently expect more than 48 hours between two consecutive reboots In these conditions the computation w
32. mpositions spec tral intervals have been considered e H0 CO2 and O3 in the 748 752 cm range e HO CO and Os in the 4900 5040 cm range e H0O CO2 Oz and C Hy in the 4200 4240 cm range In each case spectrum have been computed for 49 atmospheric levels from ground level 1 atm 292K to 2 878 10 atm 0 29 Pa 345K at the top of the atmosphere Figures 3 to 5 represent both spectrum in several spectral ranges at different pressure and temperature levels Both linear and logarithmic scales are used in order to compare highest and lowest values A 2 High pressures temperatures Absorption spectrum computed by KSPECTRUM in the case of Venus at mosphere have been used to compute absorption coefficient average ka over 68 narrowband spectral intervals covering the 40 5825 cm range These average absorption coefficients have been compared to average absorption coefficients computed from k distribution data used for radiative transfer simulation in Venus atmosphere 25 26 Comparisons have been performed at ground level 90 9atm 735K at an altitude of 5km 65 8 atm 697K and at an altitude of 55km 0 52atm 302K In each case orders of magnitude of average absorption coefficients com puted from KSPECTRUM high resolution spectrum are in good agrement with average absorption coefficients from k distribution data especially in the 1 7 and 2 3 um transparency windows this validates s
33. n This should give you the list of machines in your cluster Alternatively you can use OPENMPI instead of MPICH Installation is easier you do not have to create a mpd conf file and the MPD daemon does not have to be started mpdboot prior to code execution Program mation instructions are fully compatible 3 5 Installation of LBL databases KSPECTRUM needs to access LBL databases in order to run LBL databases are not provided with the code you will have to download them These databases can be found on FTP servers e HITRAN amp HITEMP ftp cfa ftp harvard edu e CDSD ftp ftp iao ru You should first connect to either of these FTP sites in order to locate the top level directory for the required database probably somewhere in pub If you are using MacOS it is possible to use the Go Connect to server option within the Finder Use anonymous as login and your e mail address as password Then just copy and paste the top level directory adress It is also possible to use wget if you can only use command line tools for instance if you try to access ftp sites from a remote machine the example command line below is given for HITEMP 2010 wget drc user anonymous password your_email_address ftp cfa ftp harvard edu pub HITEMP 2010 This will litteraly download the whole HITEMP 2010 directory right into the directory where this command was issued The code expects to find LBL da
34. n KSPECTRUM on you will first have to download mpich2 from http www mcs anl gov research projects mpich2 make sure you download the latest version Next untar the downloaded archive and install it on every system that will be part of your cluster configure prefix path to installation directory make make install Before running the MPD daemon you must create a your home folder echo secretword secretword gt gt mpd conf chmod 600 mpd conf using any secretword Next you will need to be able to connect via ssh to every other machine of your cluster with no password request For this you must first create a DSA key ssh keygen t dsa leaving all fields blank use the enter key to answer each question Then you will have to add this DSA key to the list of authorized keys cd ssh cat id_dsa pub gt gt authorized_keys Finally create the list of machines that belong to your cluster This list must reside within the mpd host file on your home folder Each line must contain the name of the machine by order of availability host1 domain host2 domain host3 domain 6 mpd conf file in 17 etc You can then try to run the MPD daemon mpdboot n with the number of hosts you want to run MPD on typically the number of machines in your cluster If you encounter no error you can use command mpdtrace to check the number of hosts the MPD daemon is running o
35. n is impossible e If you are low in memory and you can not increase further the value of Nkmx you can try to use a coarser spectral resolution or even a constant spectral step by setting the appropriate options in the op tions in file within the Spectral discretization algorithm section e You can also try using smaller narrowband spectral intervals Nkmx is the number of values of v a given narrowband interval will be discretized into so using smaller narrowbands can make the discretization possible e If all the above suggestion failed or you have the feeling there is an other problem feel free to contact me When possible send me your composition in narrowbands in data in and options in files with a description of the problem you encounter What about all this unknown spectroscopic data As mentioned in section 2 3 we know very little about sub lorentzian profiles collision induced absorption CIA lines and broadening by the rest of the gas mixture Hypothesis have been made sub lorentzian profiles are given in the case of pure CO and in specific spectral temperature ranges of validity These limitations can be overridden by the user see section 4 1 2 Broadening by the rest of the gas has been considered as if the rest of the gas mixture was terrestrial air and CIA can be accurately simulated for CO only in the 0 250 em spectral range only I am aware this situation
36. ne_list and OK_line_list Within each one of these directories you must create two other directories one named Compressed_files that contains original compressed data files zip files and a second one named Uncompressed _files that con tains un zipped files par files for the molecular species Upper lower case matters the code will look into this directory structure as mentionned in this section data CDSD must contain CDSD uncompressed data files If a link it must point to the CDSD_1000_ UPDATED uncompressed_data lo cal directory data CDSD HITEMP must contain CDSD HITEMP uncompressed data files data CDSD_4000 make it a link preferably to a folder that con tains every single one of the 2768 uncompressed LBL files for the 4000K version of CDSD This version of the database holds informa tion for more than 628 millions of transitions and the total weight of the database is 73 Gb once uncompressed 23 Gb for the compressed files that can be downloaded from the FTP site The files in this folder should not be renamed their names are cdsd_00220_00230 cedsd_00230_00240 cdsd_00240_00250 etc You will have to find a way to unzip every zipped file that has been downloaded using for instance a base script like this one 19 1 bin bash 2 TARGET is the folder containing target files i e zipped files 3 TARGET HOME lab_data CDSD_4000 compressed_file
37. of new transitions are activated at high temperatures The CDSD database offers a solution for CO lines it is valid up to 3000K 5 A new HITEMP database is in preparation that should cover a huge number of high temperature transitions for HO More generally the accuracy and range of application of public LBL databases is clearly increasing However the accuracy of spectra computations always depend on the available LBL parameters The current level of knowledge will always impose a limitation on the accuracy of numerical simulations WARNING there are some inconsistencies within the current HITEMP database some values of the 7se1f parameter are null This is an issue because in the case you want to produce a high resolution spectra for a single molecular species only this parameter will be taken into account in the computation of the Lorentz line width yz A null Yse y will result in a null value for yz and this will end in a crash of the code during spectral discretization The spectral discretization algorithm has been improved in order to detect null values of Yse f in this case instead of using a null value the Yser parameter will be set to the closest non null value found among the LBL database 2 4 A better respect of spectroscopy The general idea of KSPECTRUM is to ensure the simple accuracy criterion that are discussed in section 2 1 In order to achieve this for any situation gas mixture composition values of temperatu
38. of gas There fore KSPECTRUM should probably not be used for a pressure greater than 100atm The lower pressure limit will be imposed by the code itself and more specifically by the spectral discretization algorithm It may crash for values of the pressure lower than 1077 atm because spectral lines are so narrow that wavenumber steps are really small 10 7cm Besides the fact that the computation of the spectral grid may take a lifetime in these conditions such small wavenumber steps may not be properly taken into account by the algorithm In terms of molecules the HITRAN database takes into account the tran sitions of 42 for the 2008 version molecular species along with their iso topes The list of molecules can be found in the HITRAN documentation 1 or directly in the data molparam txt file Other limits may be imposed for specific gas mixtures See section 2 3 for more details 2 3 Physical hypothesis This section presents how we manage the fact that there is a lack of available spectroscopic data for various aspects of the theory 2 3 1 Sub lorentzian profiles and CIA The main issue is that we have very little knowledge about e The real line shape for each molecule at each wavenumber We know that CO lines have a sub lorentzian behavior in line wings which is why thermal infrared signals can be transmitted from ground level up to space in very specific near IR spectral windows through the otherwise very thi
39. omputed the code knows the number of k values that will have to be computed The computational load that will be sent to each process is the number of k values divided by the number of chunks If this number of chunks is large each process will have to compute a small number of k values A chunk will be sent each time a process is found idle The advantage is that if processors run at different speeds have to run other processes processes that run on slow processors will be sent a lower computational load fewer chunks and processes that run of fast processors will be sent a greater number of chunks e Section Levels and narrowbands limits first the high temperature level has to be indicated this value is used to switch between HITRAN at low temperatures and HITEMP or CDSD at high temperatures By using a high value higher than the highest value of temperature 22 you want to compute spectrum for you can be sure only HITRAN is used The user can then specify the indexes of the first and last atmospheric level narrowband interval if only a limited number of atmospheric levels narrowband intervals should be treated instead of every atmospheric level narrowband interval The choice itself is done via the corresponding option e Section Special options for degraded mode the user should pro vide here values the code will use when running into degraded mode Three values are used for the degraded spectral discretizati
40. on algorithm number of line widths to consider in the definition of a line s central zone the number of points used to discretize lines central zones and the number of points used to discretize regions between lines The next parameter is the minimal distance from current value of v to the central wavenumber of each transition that is used for line profiles truncation Then the user should provide the percentage of transitions in terms of reference line intensities that will be neglected by the reference spectral discretization algorithm One last value is the value of ce used by the convergent k algo rithm see options in when this algorithm is enabled transitions are sorted according to their intensities and the contribution of most intense lines are added first when computing k v The algorithm will regularly update and check the evolution of k v this history is expressed as successive values of a me with k v the succes sive values of k v that have been recorded and k v the current value of k v the last recorded value Convergence is reached once Qn An 1 lt Max a aizi 1 e With the maximum error per centage allowed over k v and ec a value that provides the user a mean to control this convergence criterion 4 1 2 The options in file e Section Narrowband discretization the user can chose between com puting automatically the narrowband discretization or use pre
41. r the 2004 version of HITRAN if you want to use it you can select which database you want to use by set ting the right option in the options in file see section 4 1 2 and for the HITEMP and CDSD databases 20 Frequent updates are published about HITRAN data If you want to use these new data files follow these directions let s say you want to download the update for SO from the HITRAN website section HITRAN updates Download the updated file O9_hit0O9 par within the folder that contains your by molecule data files KSPECTRUM will still look for files that are labelled 08 so you will have to link the new file to the old one within the by molecule data folder rename the old SO file mv 09_hit08 par 09_hit08 par ori Then create a link from the new file to the old one In s 09_hit09 par 09_hit08 par Repeat this step for every update file Finally there is also supplemental data that is not directly provided among the by molecule data These files reside within the HITRAN2008 Supplemental folder There are currently three data files within this supplemental folder 30_hitO8 par 35_hit08 par and 42 hit08 par In order to add then to your by molecule folder you can just copy paste them or once again create symbolic links does not take more disk space within your by molecule data folder type the following commands In s Supplemental 30_hit08 par
42. re and pressure it is absolutely necessary that the coding is well separated from the physical problem In particular the available computing power should not impose limitations on the versatility of the code We believe this is possible with to days mass computing power 2 4 1 no line selection Existing computation codes that can produce high resolution spectrum al ways involve line selection at some point The purpose of selecting lines is to reduce the total computation time by reducing the number of lines N whose contribution has to be computed at each wavenumber point v see Eq 1 However KSPECTRUM does not need to select lines that have to be taken into account the contribution of all transitions are effectively com puted See section 2 1 for details accuracy parameter 11 2 4 2 no truncation of lines and continuums In the same manner line profiles are not truncated as they probably would in a classical approach the contribution of each line at each wavenumber v is computed no matter how far from the line center v is This is also a direct consequence of the way accuracy parameter is taken into consideration KSPECTRUM does therefore not need to use a continuum that is clas sically added to the computed spectrum to correct line truncation side effects Values of k v are computed with a specified relative accuracy 1 2008 12 09 update this is no longer completely true since I have en ab
43. s 4 RESULT is the folder containing resulting files i e unzipped files 5 RESULT HOME lab_data CDSD_4000 uncompressed_files 6 7 if d TARGET s then 9 echo TARGET folder does not exist 10 echo TARGET 11 exit 1 i fi ig if d RESULT 14 then 15 echo RESULT folder does not exist 16 echo RESULT 17 exit 1 is fi 20 for FILE in TARGET x zip 21 do 22 echo Now unzipping file FILE 23 unzip FILE d RESULT 24 done 25 exit 0 In practice let s say you have downloaded and unzipped the 2008 ver sion of the HITRAN database somewhere on the local disk of the machine you want to use for your computation KSPECTRUM will need the by molecule data files that are provided with HITRAN For instance you have these files in the following folder HOME LBL_databases HITRAN HITRAN2008 By Molecule Uncompressed files On the other hand KSPECTRUM is installed in the following folder HOME great_codes kspectrum1 2 0 In order to let KSPECTRUM know where the by molecule data files of the HITRAN2008 database are located go to the data folder cd SHOME great_codes kspectrum1 2 0 data and create a symbolic link called HITRAN2008 to the folder that con tains the data files In s SHOME LBL_databases HITRAN HITRAN2008 By Molecule Uncompres files HITRAN2008 Here you go You have just installed the HITRAN2008 database in KSPECTRUM You can do the same fo
44. s and M J Linevsky Integrated intensities of No CO 2 and S Fs vibrational bands from 1800 to 5000 cm Journal of Quantitative Spectroscopy and Radiative Transfer 42 465 476 1989 M Gruszka and A Borysow Roto translational collision induced ab sorption of CO for the atmosphere of Venus at frequencies from 0 to 250 em and at temperature from 200k to 800k Icarus 129 172 177 1997 17 18 19 20 21 23 24 25 26 39 M Gruszka and A Borysow Computer simulation of the far infrared collision induced absorption spectra of gaseous C O2 Molecular physics 93 No 6 1007 1016 1998 J B Pollack O B Toon and R Boese Greenhouse models of Venus high surface temperature as constrained by Pioneer Venus measure ments Journal of Geophysical Research 85 8223 8231 1980 Y Baranov W J Lafferty and G T Fraser Infrared spectrum of the continuum and dimer absorption in the vicinity of the Oz vibrational fundamental in O2 CO 2 mixtures Journal of Molecular Spectroscopy 228 432 440 2004 P Dubuisson J C Buriez and Y Fouquart High spectral resolution solar radiative transfer in absorbing and scattering media application to the satellite simulation Journal of Quantitative Spectroscopy and Radiative Transfer 55 103 126 1996 P Dubuisson R Borde C Schmechtig and R Santer Surface pressure estimates from satellite data in the oxygen A band Applications to the MOS
45. s that the size of each array has to be declared And in the present case the most obvious limitation is the size of the arrays that hold the LBL data For a typical atmosphere Earth Venus less than 10 molecular species have to be taken into account which gives a total num ber of lines of approximately 2 10 when using the HITRAN LBL database 12 However when using the CDSD database for CO lines the number of lines that have to be taken into account can exceed 107 And there is no way 9 ar rays of size 10 can be declared on a classical machine that holds 1GB RAM KSPECTRUM actually needs to read 9 LBL parameters from databases which imposes the declaration of 9 data arrays The default size of the ses arrays is 2 10 and can be reduced for low memory configurations see section 3 2 With such a limitation the only way several millions of lines can be taken into account is to perform a multi pass computation for instance when the user specifies the use of the CDSD database for CO transitions 10 millions lines have to be taken into account for CO only With an array size of 10 KSPECTRUM will read the first 10 lines first pass and perform the computation of the k v spectrum with these first set of lines it will actually compute the contribution of this first million of lines to the final result Then it will perform a second pass using the second million of lines in the database it will compute the contribution
46. se validity ranges Options see section 4 allow the use of CIA computation outside these validity ranges In particular the 2 3 and 4 7um spectral bands of CO are not taken into consideration 2 3 2 Broadening by the rest of the gas Another limitation of the code is the hypothesis that is done concerning Yother lines of a given species 7 can be broadened by collisions with the same species i or by collisions with other species j The self broadened half width 7se1 given by LBL databases and that account for collisions between molecules of the same species can be considered as reliable However LBL databases also give us parameter Yair the air broadened half width that corresponds to the broadening of lines by collisions between species i and the air on the terrestrial atmosphere Of course when computing a spectrum for a non terrestrial atmosphere parameter Yother Should not take the value of Yair indicated in LBL databases because the rest of the gas is no longer Earth air However we have no other choice than taking Yother Yair because there are no other data available This choice can be justified by the fact that Yorn will probably never be very different from Yair anyway Note that M Gruszka s code is not supposed to be redistributed However the authors mention that the code can be used in academic applications such as KSPECTRUM 10 2 3 3 High temperature transitions We know that a large number
47. sensor over land Journal of Geophysical Research D 106 27277 27286 2001 P Dubuisson R Borde D Dessailly and R Santer In flight spec tral calibration of the oxygen A band channel of MERIS International Journal of Remote Sensing 24 No 5 1177 1182 2003 P Dubuisson D Dessailly M Vesperini and R Frouin Water va por retrieval over ocean using near IR imagery Journal of Geophysical Research 109 D19106 2004 P Dubuisson V Giraud O Chomette H Chepfer and J Pelon Fast radiative transfer modeling for infrared imaging radiometry Journal of Quantitative Spectroscopy and Radiative Transfer 95 201 210 2005 M A Bullock and D H Grinspoon The recent evolution of climate on Venus Icarus 150 19 87 2001 V Eymet R Fournier J L Dufresne S Lebonnois F Hourdin and M A Bullock Net Exchange parameterization of thermal infrared ra diative transfer in Venus atmosphere Journal of Geophysical Research 114 E11008 DOI 10 1029 2008JE003276 2009
48. spheric level with the highest pressure level Each file contains 3 columns The fist column gives values of wavenum bers inem The second column is the total cross section of the gas mixture in cm molecule at the corresponding wavenumber The third column is the total absorption coefficient in m t at the corresponding wavenumber 5 2 More output files Additionally the user can check a number of log files produced by KSPEC TRUM during its execution e results composition_info txt information returned by the algorithm that reads the atmospheric composition file and sets abundances for each molecular species This file shows what molecules have been iden tified the number of isotopes and their respective abundances For each molecular species the sum of abundances of all the isotopes should be equal to 1 e results calculation_info txt various output information values of accuracy criterion and are given along with the gain from line classification the total number of lines is printed as well as the num ber of lines whose contribution has to be explicitly computed at each value of v These are cumulated values over each narrowband inter val over each atmospheric level The difference is an estimation of the computation time gained by line classification using criterion 1 These two files should be saved together with spectrum result files in order to keep a record In particular file results
49. tabases in the data directory Of course you have to download LBL databases in a directory that is common to all machines the code will run on shared disk so that each process will be able to access the LBL databases through the data directory LBL data file directories must be the following e data HITRAN2004 must contain the uncompressed by molecule HITRAN data files in its 2004 version If you create a link it must 18 point to the HITRAN2004 By Molecule Uncompressed files direc tory on the local disk of each machine When the code is set to use the HITRAN2004 database see options in in the 4 1 2 section this folder must exist within the data folder data HITRAN2008 must contain the uncompressed by molecule HITRAN data files in its 2008 version If you create a link it must point to the HITRAN2008 By Molecule Uncompressed files direc tory on the local disk of each machine When the code is set to use the HITRAN2008 database see options in in the 4 1 2 section this folder must exist within the data folder data HITEMP must contain the old HITEMP database uncom pressed data files If a link it must point to the HITEMP uncom pressed_data local directory data HITEMP 2010 must contain the new 2010 version HITEMP database Within this folder you must have 5 folders named CO2_line_list CO_line_list H2O_line_list NO_li
50. ub lorentzian profiles representation and in the far infrared v 300 em which vali dates the collision induced absorption continuum 33 Table 1 provides a comparison between average absorption coefficients computed at ground level 0 0016 34 0 01 LOA 010 LOA 010 0 0014 kspectrum 010 kspectrum 010 0 0012 0 001 0 001 H 0 0008 0 0001 a x 0 0006 0 0004 16 05 0 0002 0 1e 06 f f f f f f 748 748 5 749 749 5 750 750 5 751 751 5 752 748 748 5 749 749 5 750 750 5 751 751 5 752 v cm v cm a b Figure 3 Absorption spectra for a H20 CO and O3 mixture at pressure of 0 298 atm zm o 3 34 1074 rco 3 30 1074 zo 1 20 1077 and a temperature of 238 6K in the 748 752 cm spectral range 0 006 LOA 005 0 1 LOA 005 kspectrum 005 i kspectrum 005 0 005 0 01 1 0 004 0 001 0 003 E 7 0 0001 0 002 J 0 001 19 09 i o AM Jul Wa wud 1e 06 1 1 I 4940 4945 4950 4955 4960 4965 4970 4975 4980 4940 4945 4950 4955 4960 4965 4970 4975 4980 v cm v cm a b Figure 4 Absorption spectra for a H20 CO and O3 mixture at pressure of 0 583 atm zmo 3 04 10 3 zco 3 30 1074 xo 5 16 1078 and a temperature of 270 3K in the 4940 4980 cm spectral range x 35 00014 l LOAO ca n L
51. usin R LeDoucen C Boulet and A Henry Temperature depen dance of the absorption in the region beyond the 4 3 um band head of CO3 2 No and O broadening Applied Optics 24 No 22 3899 3907 1985 V Menoux R LeDoucen and C Boulet Line shape in the low frequency wing of self broadened CO lines Applied Optics 26 No 3 554 562 1987 M Y Perrin and J M Hartmann Temperature dependant measure ments and modeling of absorption by CO N mixtures in the far line wings of the 4 3m CO 2 band Journal of Quantitative Spectroscopy and Radiative Transfer 42 311 317 1989 M V Tonkov N N Filipov V V Bertsev J P Bouanich Nguyen Van Thanh C Brodbeck J M Hartmann C Boulet F Thibault and R LeDoucen Measurements and empirical modeling of pure CO 2 ab sorption in the 2 3 m region at room temperature far wings allowed and collision induced bands Applied Optics 35 No 24 4863 4870 1996 N I Moskalenko Y A Ilin S N Parzhin and L V Rodionov Pressure induced IR radiation absorption in atmospheres Izvestiya Atmospheric and Oceanic Physics 15 632 637 1979 T G Adiks Absorption spectrum of CO in the 2 15 4 um region under conditions of high pressure and CO content Optics and Spectroscopy 40 375 378 1976 T G Adiks Influence of the state of agregation of CO on the intensities of allowed and induced absorption bands in the 1 4 um region Optics and Spectroscopy 44 541 544 1978 M E Thoma
Download Pdf Manuals
Related Search