Indian Journal of Radio & Space PhysicsVol. 21. February 1992. pp. 26-32

f\Am~gorithm to derive vertical profiles of atmospheric gases using

solar spectra_._._...~_ _. \.~. ·~r·.-

, ~ Mehra &b ~ Jadhav "\( Indian Institute of Tropical Meteorology, Pune 411 008 )

Received 3 December 1990; revised received 17 June 1991•. A method is described for retrieving the vertical profiles of the atmospheric gases using ground basedspectroscopic observations of solar spectra in the visible region. The algorithm described assumes thatthe vertical profile of the species under consideration does not vary during the period of observation, andthe absorption cross-sections of the species do not vary with pressure. There is good comparison be-tween the assumed profiles and the model computed profiles. The maximum number density layer is rep-roduced at the same height as in the assumed profile, and any fine structure in the assumed profile is alsoresolved. The five layer'S assumed for analysis show a ± 10% difference between the assumed profile andthe computed profile, whereas, this difference is ± 5% for the nine assumed layers. The observations car-ried out with visible spectrometer are used to derive vertical profile of ozone for nine assumed layers.The comparison with Umkehr data shows the deviation of the order of ± 15%. '

I IntroductionThe knowledge of atmospheric activities is im-

portant from the point of view of suitability of theplanet Earth for mankind. The day-to-day activities,direct or indirect, are disturbing the useful ozonelayer, which is acting as a shield to protect the life onthe Earth from the harmful ultraviolet radiations.The ozone depletion, greenhouse effect, sea levelrise etc., are some of the current problems. Themonitoring of atmospheric ozone and other atmos-pheric constituents is the major requirement of to-day. A versatile system (Uv-visible spectrometer)has been developed at the Indian Institute of Tropi-cal Meteorology (IITM), Pune, which may give thevertical profiles of the atmospheric constituentswith a lesser complicated process as compared tothat used for the in situ measurements using balloonand aircraft borne instruments. The instrument hasbeen described in detail in an earlier paper'. How-ever, for the sake of completion, Fig. 1 gives theblock diagram of the instrument. Various other de-tails and specifications of the instrument are given inTable 1. The preliminary results of the observedozone vertical profile are also presented.

In this paper. the procedure for an algorithm toretrieve the vertical profiles of different gases usingground based spectroscopic observations of the so-lar spectra in the visible region is described. Thetheoretical calculations for testing the algorithm andthe accuracy of the procedure, and the assumed and

computed vertical profiles of NOz and 03 are pre-sented.

2 TheoryIt is possible to derive the column density of an

absorbing gas species by taking observations of so-lar spectra at different solar zenith angles in the re-gion of absorption by the species. If the number ofabsorbing species is more than one in a particularspectral region, then it is possible to separate outthese species by technique like matrix inversion.

The following assumptions have been made whiledeveloping the theory for the retrieval of vertical

ENT-SLIT

RE CORDER

SCANNIN. IYS]CONT"OL

MULTICHANNEL

AVERAGER AND

Fig. I-Block diagram of high resolution visible spectrometerfor atmospheric studies

MEHRA & JADHAY: ALGORITHM FOR VERTICAL PROFILES OF ATMOSPHERIC GASES 27

Parameter

Table 1- Detail specifications of high resolution visible spectrometer

Specification

Czerny-Turner-Ebert-Fastie (CTEF)Size=128xl54mmGrooves per mm = 1200Blaze angle = 460

Blaze wavelenth = 1.2 utn in 1st order(This grating is to be used in Hnd to IVth order)

1500 mm with mirror diameter 220 mm3.85 Nmm in 1st order1.93 Nmm in Ilnd order1.30 Nmm in IIIrd order0.96 Nmm in IVth order

0.190 A in 1st order0.100 A in IInd order0.065 A in IIIrd order0.050 A in IVth order

50 Jlm to 10 mm150mm150mmSquare-wave type transmission filters by Corion Corporation, USA.The periodic rotation of grating through a scanning nut and screw arrangement. Thescrew is rotated by a stepper motor whose direction is controlled by a personal compu-ter. The data is stored on floppy for further processing.Hamamatsu photomultiplier Tube R763 with spectral response 5-500 (better than5-20). Cooling system by Thorn E.M.I. can cool the system up to - 200C

Type of the systemPlane diffraction grating

Focal length of the systemDispersion of the system

Maximum resolution (at 50 Jlm slit width)

Opening of the adjustable curved slitRadius of curvature of slitsHeight of the slitsOrder sorting filterScanning system

Photomultiplier tube with cooling system

profiles of different gases:(i) The vertical profile of the species under consid-

eration does not vary during the period of ob-servation.

(ii) The absorption cross-sections of the absorbingspecies do not vary with altitude, i.e. with pres-sure.

(iii) The direct solar spectra observations are to betaken for different solar zenith angles. Thenumber of observations to be taken dependsupon the number of layers to be observed.

2.1 Method ofanalysisTo begin with, a five-layer model of atmosphere is

assumed for the study. The details of the procedureare given below:

The attenuation of monochromatic radiation, ac-cording to Lambert-Beer's law, is related to thenumber of absorbing molecules in the optical pathso that,/= /0 exp (- aMX) ... (1)where,/ = Measured flux/0 = Incident flux at the top of the atmospherea = Absorption cross-section of the absorber for a

particular wavelengthM = Density of the absorbing gas in the pathX = Path length, i.e. optical path traversed by solar

radiation, and it is a function of altitude andzenith angle.

However, the attenuation of the monochromaticradiation is not only due to the absorber gas, but al-so due to Rayleigh and Mie scattering and the'Fraunhofer filling in' (FFI) effect. The Lambert-Beer's law with these parameters can be written as

/= /oexp( - asMsX- aabMabX- amsMmsX) ... (2)where,as and ams Rayleigh and Mie scattering. cross-sec-

tions, respectively, for a particular wave-length

aab Absorption cross-section of the absor-ber for a particular wavelength

M, Density of air molecules in the pathMao Density of absorbing gas in the pathMms Density of aerosols in the path

The effect due to Rayleigh and Mie scattering andthe FFI should be removed before using the algo-rithm for the retrieval of the vertical profile of theabsorber gas. The detailed procedure for retrieving

28 INDIAN JRADIO & SPACE PHYS, FEBRUARY 1992

the slant columnar density of ozone is explained inSec. 3. For calculating the effect of Rayleigh and Miescattering and FFI, the method given by Solomon?can be utilized. In this method, as in all absorptionmeasurements, the ratio of any particular spectrumwith the corresponding background spectrumshould be the one obtained outside the Earth's at-mosphere, so that it is entirely free from atmosphericabsorption. However, the ratio of spectrum taken ata small solar zenith angle (i.e., around noontime)and large zenith angle (i.e. eveniogtimes) can elimin-ate the background spectrum outside the atmos-phere", The spectra ratio allows measurement onlyof the change in the slant column relative to thebackground spectrum. This is useful in the detectionof weak features in the observed spectrum':", Thus,use of the spectra ratio in the analysis eliminates thenecessity of computing the extra-terrestrial value ofthe intensity (10) at the wavelength used for thestudy. The spectra ratio is thus derived from theevening and noontime observations. A linear leastsquares method is used to analyze this spectra ratiowhich contains information regarding Rayleigh andMie scattering, the absorption by the gas species andthe FFI effect. Using the inversion procedure, theabove atmospheric parameters can be retrieved+'.Thus, using Lambert-Beer's law, the attenuation ofthe light intensity reaching at the ground due to Ray-leigh and Mie scatterings and the FFI effect is sepa-rated. The number density of the absorption gasspecies traversed by solar radiation as computed bythe inversion method is used to retrieve the verticalprofile of the absorber gas species.

The retrieval of the vertical profile of the atmos-pheric gases can be carried out from the slant co-lumn densities of these species at different solar zen-ith angles. The total set of calculations is presentedfor 5 assumed layers and also extrapolated for 9 lay-ers. However, the same logic can be used for highernumber of layers. The contribution due to N02

above50kmis ""1.0%andthatof03is ""0.1-1%of their total column densities. Hence, in the presentcalculation, the contributions from above 50 km arenot incorporated. However, on similar lines thesame can be incorporated in the retrieval of densit-ies at higher altitudes. In this study, the total heightof the atmosphere is assumed to be 50 km. Each ofthe five assumed layers is considered to be of 10 kmthickness. Thus, using Lambert-Beer's law we get

III = IOJ exp] - al (mlxl + m2x2 + m3x3 + m4x4+ m5x5)]

and

1).2 = I02exp [- a2 (mixI + m2x2 + m3x3 + m4x4+ msxs)]

where,

Ill' 1).2 Measured flux at the ground for wave-lengths Al and A2, respectivelyIncident flux at the top of the atmo-sphere for wavelengths Al and A2, res-pectivelyAbsorption cross-sections of the absor-ber gas at wavelengths Al and A2, re-spectively

ml, m2 •• ··mS Number-of the densities absorbinggas in each consecutive 10 km layerof the atmosphere

XI' x2, ",X5 Optical path lengths traversed by solarradiation for a particular zenith anglein the 5 consecutive layers of thicknesslOkmeach

The path lengths Xl' X2, •.• Xs can be calculated by us-ing the geometry shown in Fig. 2 (Ref. 2).

The ray paths are calculated for solar zenithangles, and the solar zenith angle x changes accord-ing to the layer altitude.

By taking ratio of the two adjacent wavelengths attime 11 the following equation can be written fornoontime observation

ZENITH

III10

I II II ~R(\ II I\III'I,

FOR 'X. ~ 90°0- RE +Z, b- RE + Z+ IlZ

a: = sin-I[~ .sin"]

ex - [2 2 OoS- a +b-2abcos(x-a:>JFig. 2-Schematic diagram of viewing geometry of the sun andthe trigonometric relations used for the calculation of slant pathlengths (Solomon et al. 1987) [RE = Earth radius, Z= Altitudeof the layer assumed (lower side), x= Solar zenith angle, b= Al-

titude of layer assumed (upper side)]

MEHRA & JADHAV; ALGORITHM FOR VERTICAL PROFILES OF ATMOSPHERIC GASES 29

- - msx~'(al - a2)) '" (3)

where, x;' = Path length in the ith layer at time II'and r= 1,2, ...5.The ratio of the intensities at the two adjacent wave-lengths at time 12, i.e. for any other solar zenithangle, can be written as

- - msx~2(al - a2)]

Taking ratio of Eqs (4) and (3), we have... (4)

R'2RR = --.!!= exp[ - m (X'2 - x")(a - a )- .21 RI, I I I I 2

12

-ms(x~2-x~')(al-a2)] ... (5)The intensities outside the atmosphere at A I and Az,i.e. 10l and 102 are thus eliminated.Taking the ratio of time 13 and 11' we have

R"RR =--.!!=exp[-m (xlJ-x")(a -a )_ .31 Rh I I I I 212

. . . (6)

Similarly ratios of time 14and II, ts and tl and t6 andII can be computed. Linearizing the above equa-tions, the following five equations can be writtenIn(RRzI) = - ml(x~2 - x~')(al - az) - .

- m5(x~'- x~')(al - a2) ... (7)In(RR31) = - ml(x;'- x~')(al - az) - •.•..•

- m5(x~' - x~')(al - az) ... (8)

In(RR41) = - ml.(x~' - x~')(al - az) - •.••••

- m5(x~'-x~')(al- az) ... (9)In(RRs1)= - ml(x'j-x't)(al- az)- ..•...

- ms(x~'- x~')(al - az) ... (10)In(RR61) = - ml(x;' - x;')(al - az) - •..•••

- ms(x~'- x~')(al- az) ... (11)In Eqs (7 )-( 11), the difference in number of mole-

cules between noontime and eveningtime observa-tions can be used.

In RHS of Eqs (7)-(11), (al - a2) is common andthe remaining factor is nothing but the product ofmolecular density and the distance through which

the ray is traversed, i.e. the remaining factor is theslant column density.

Let MI21, be the observed difference in numberof molecules between the time of observations 12and 11' Eqs (7H11) can be rewritten as- M'2I'(al - az) = - m\{x': - x';)(aJ - a2)-· •••••

- ms (X~2- x~,)(al - az) ... (12)

- MI,/'(al - az) = - mdx;' - x;')(al - az) - ••••••

- ms(xi- x~')(al- az) ... (13)

- u=;«, - az) = - ml (x;' - x;')(al - az) - ••••••

- ms(x~'- x~')(al- a2) .... (14)- M""(al - a2) = - ml (x;' - x;')(al - az) - ••••••

- ms(x~- xi)(al - az) •.. (15)- M'~"(al- az)= - ml (x;'- x;')(al- az)- ••••••

- ms(.~~·- x~')(al - az) ... (16)In this manner, five independent equations are

obtained. In Eqs (12)-(16), al and a2 are knownand the optical path lengths through various layersare also known (calculated geometrically). Thus,there are five unknown quantities in Eqs (12)-(16)i.e. ml, mz, •.. ms' The above linearized equations[Eqs (12)-(16)] can be written in the matrix form asfollows.[In(RR2,) In(RR3,) In(RR,,) In(RR,,) In(RR.,ll=

[m, m2m3m,m,l [- (x;' - x;')(a, - a2)····· - (x;' - x';)(a, - a2/]- (x;' - x;')( a, - a2)····· - (x;' - x;')( a, - a2).-(xj'-x;')(a,- a2)····· -(x~- x;')(a,- a2)- (xi - x~)(a, - a2)····· - (x; - x~')(a, - a2)- (x~'- x~')(a, - a2)····· - (x;' - x;')(a, - a2)

or,[M"" M'l" M"" MI,t, M""] =

[m m m m m] (X'I' - x'I')··· ··(X'l' - Xii')I Z 3 4 4

(X;2- x;')· ...• (x;' - x;')

(X;2- x;')···· ·(x~ - x;')

(x~ - x~')···· ·(xi - x~')(X~, - x~,).•... (x~·- x~')

This can be represented in the formA=BXAX-l = BXX-rAX-l= BI

Thus from the elements of the matrix AX - I, theelements of the matrix B can be calculated. The ele-ments of matrix B gives the vertical profile of the ab-sorber gas, i.e. the values of m1, m2, ••• m..

For testing the above algorithm and to determinethe percentage error of the procedure the following

30 INDIAN J RADIO & SPACE PHYS, FEBRUA.RY 1992

calculations are carried out. The mean total columndensity of 03 (or N02) and its mean vertical profileare assumed for these computations. The attenuatedintensity due to the assumed number density of theabsorber species under consideration is computedfor different solar zenith angles. The intensities thuscalculated are used in Eqs (3)-(6) to re-calculate thenumber density vertical profiles of 03 and N02 se-parately. The details of the wavelengths used for re-trieving 03 (or N02) are given in Table 2. The wave-lengths used for these calculations are having maxi-mum difference for the spectral region considered.The absorption cross-sections at the above-men-tioned wavelengths were taken from Refs 5 and 6.The fine structures assumed in the profile can be re-solved in the computed vertical profiles of N02 and03. The five-layer model computations show± 10% difference between the assumed profile andthe model computed profile as shown in Figs 3 and 4.

Table 2-Wavelengths used for model calculations

Model Wavelength pair used for

N02 03

A A5-layers 4485;4405 4()OO;6000

9-layers 4485;4405 4000;60004398;4381 4374;5900

-- ASSUMED VERTICAL PROFILE

---- MODEL PREDICTED VERTICALPROFILE~~~~~ t::

~20~ ~

I ~O I 2 3 4 5 6 0 I 2 3 4 5 6 01 2 4

O. DENSITY, X 10'2_lecule/c •• 1

Fig. 3-Comparison between the assumed and model-computed5-layer vertical profile for 03 (Three different structures of thevertical profile of 03 have been assumed and compared with the

computed profiles)

ASSUMED VERTICAL PROFILEMODEL PREDICTED VERTICALPROFILE

Fig. 4-Same as Fig. 3 but for NO!

For retrieving the vertical profile of the species inthe nine different layers of 5 km thickness each,there are 9 unknowns; hence, at least 9 independentequations are required to obtain the vertical profileof the species. The 9 independent equations can beobtained by following three approaches, i.e.

(i) Using intensity ratios of observations at twowavelengths at 5 different zenith angles with thecombination of different zenith angle observa-tions

(ii) Using intensity ratios of two wavelengths at 10different zenith angles

(iii) Using observations at five different zenithangles for two different pairs of wavelengths(i.e. four wavelengths)

If the first method is followed, spurious results areobtained. This may be due to the fact that the sameobservations are repeatedly used and this causes thesolution of the matrix to oscillate, thus giving spuriousresults. However, the second and third method givea good comparison between the assumed profileand computed profile. The above results are testedby assuming different structures of the vertical pro-files. The computations show that the maximumnumber density layer is reproduced at the sameheight as in the assumed profile, and fine structurein the assumed profile is also resolved. The nine as-sumed layers computation shows ~ 5% differencebetween the assumed profile and the model com-puted profile as shown in Figs 5 and 6.

-- ASSUMED VERTICAL PROFILE------ MODEL PREDICTED VERTICAL PRO"LE

~bb~-:--7--t--:'4 0 I 2 3 0 I 2 3

03 DENSITY, x 1012 mole-cule- / em3

Fig. 5-Same as Fig. 3 but for a nine-layer vertical profile of 03

ASSUMED VERTICAL PROFILE

MODEL PREDICTED VERTICAL PROFILE

E...•....:z:CDW:z:

0·51·0

Fig. 6-Same as Fig. 5 but for N02

MEHRA & JADHAV: ALGORITHM FOR VERTICAL PROFILES OF ATMOSPHERIC GASES 31

3 Experimental validation of algorithm forderiving the vertical profile of ozone

For deriving the 03 vertical profile (9 layers), theversatile UV-visible spectrometer at the Institute(specification given in Table 1) is used for the ob-servations of the solar spectra in the wavelengthrange 5460-6230A with 8A spectral resolution.The time required for scanning this spectral range isapproximately lOs. During single observational cy-cle, two spectra are recorded-one during forwardmotion of the scanning screw and the other duringreverse motion of the scanning screw. The experi-mental set-up is shown in Fig. 1. A plane mirror ofdimension 30 em x 60 em is used to direct solar ra-diations on MgO screen, placed in field of view ofthe spectrometer. The field of view of the spec-trometer is 0.08 x 5.7 deg". The spectrometer's fieldof view is completely filled by MgO screen and thesolar beam can remain on spectrometer for 8-14min. Due to this observational arrangement the track-ing vibrations are eliminated and the scintillation ef-fect is also averaged considerably. On our observedspectra, very small effect ("" ± 0.5%) is observed dueto scintillation using this observational set-up. Thesolar spectrum measurements are taken during noon-time and eveningtime. The scattered sky spectracontribution is observed during observation set bydiverting the solar beam away from MgO screen.During sky spectra observation other parameterssuch as photomultiplier supply voltage, gain of theamplifier etc. are kept constant. These spectra aresubtracted from the earlier solar spectra to applycorrection for scattered radiations. The correctedsolar spectra are used for further processing. Thespectra ratio of eveningtime sun and noontime sunis used to derive slant column density of the species.The spectra ratio is smoothed with eye estimation toremove FFI effect. The method described by $0-lomon? is used to separate out the contribution dueto different processes. Since water vapour is alsohaving strong absorption bands in this spectral re-gion, a critical examination of the spectral region wasdone and six wavelengths were chosen for ozone ab-sorption so as to avoid the inference due to the wa-ter vapour; the wavelengths used for the analysis are5590,5635,5858,6080,6145 and 6185A. The ab-sorption cross-sections are used for correspondingwavelengths in Eq. (2) for deriving the contributiondue to Rayleigh and Mie scattering processes. Thus,the slant column density of ozone is derived fromspectra ratio. The slant column densities during 10different solar zenith angles are used to retrieve the

vertical profile of the absorber gas species accord-ing to the algorithm described in this paper.

A typical afternoon observational scan in thespectral region of interest, evening observationalscan and the ratio of eveningtime sun spectrum andafternoon sun spectrum are shown in Fig. 7. Solarspectra observations were taken during 17-18 Feb.1991 and retrieved vertical profile of 03 is com-pared with Umkehr profile. Since the Umkehr pro-files for the above two days were not available, theretrieved vertical profiles are compared with aver-age Umkehr profiles for February 1990. The com-parisons are shown in Fig. 8. It is seen that there is adifference of ± 15% between the two methods.However, the general shape of the compared pro-files matches well with each other. This confirms thevalidation of algorithm. However, more observ-ations during the same period are required for theircomparison and estimation of accuracy of the algo-rithm. Our algorithm introduces error of ± 5% inretrieved profile and other error between the twomethods may be due to our comparing the average

Io-r---r---,--r-----r--r---r--r--.-------r---, 1·10

9

8

18:19:19 18:19:49AII 'E •. " ••

12:26:11012:27:21

~ 7".e 6.•

1·00

0-90o...C

0'8011:~ 5

= 4•••~ 3

2 0·70

o 0'6054tH) ~7 561·4 569·1 576·8584·5 592·2599-9607"6615-3 623'0

SPECTRAL REGION, nm

Fig. 7-0bservations of solar spectra during noontime (B) andeveningtime (A) and their ratio (NB) in spectral region 546.0-

623.0 nm on 18 Feb. 1991

17-2-91UMKHER PROFILE

50 -- II T M -RETRIVEDPROFILE

18- 2-91

...J

lO" 10'2 10" 1012

OZONE N UMBER DENSITY, molecule I cm3

Fig. 8-Comparison of vertical profile of ozone by Umkehrmethod and the algorithm developed at IITM, Pune

32 INDIAN J RADIO & SPACE PHYS, FEBRUARY 1992

Umkehr profile with daily profile of ozone during17-18 Feb. 1991.

4 Limitations of the methodThe limitation of this method, however, is dictat-

ed by the accuracy of the measuring system and al-gorithm used. The time interval between the observ-ations is decided by the amount of expected absorp-tion at the time of observation. For present observ-ations the measurement accuracy of spectra ratio is± 0.5%. The observed percentage depth for the ra-tioed spectra is from 0 to 25%. The observations foranalysis are done from 5 to 25% depth in spectra ra-tio, i.e. the observational error in the system is from± 2% to ± 10%. The total error in the retrieval ofvertical profile for nine assumed layers is the effectof measurement error ( ± 2 to ± 10%) and algorithmerror ( ± 5%). The addition of these two errors cangive ± 7 to ± 15% error in the retrieved profile. Theobserved error for the' above profile is ± 15% ascompared to Umkehr derived vertical profiles.

The N02 and 03 observations in the spectral re-gion 4360-4500A can give maximum absorption ofthe order 4-5% (Ref. 4). The percentage absorptionduring noontime is of the order of 0.1 % and in-creases with the increase in zenith angle. Hence, theobservations should be taken at 0.1, 1,2, 3,4 and5% absorptions. Thus observations at noontime,and further at zenith angles 85, 87, 88, 89 and 90°for the above expected absorptions are required.For, these type of observations with our ± 0.5%measurement accuracy can lead to ± 10% to ± 50%error due to measurement accuracy only. Hence,this spectral region is not selected for testing thepresent algorithm.

5 SummaryThe results of the above study can be summarized

as follows.(i) By taking observations of solar spectra, it is

possible to retrieve the contribution of differentspecies, which can be utilized to derive the ver-tical profiles of these species.

(ii) The algorithm used for five-layer model, i.e. us-ing two wavelengths at different optical absorp-tion cross-sections, gives the retrieved verticalprofile with ± 10% accuracy for N02 and 03,

(ill) For nine assumed layers with four differentwavelengths or ten different observations at tendifferent zenith angles gives ± 5% accurate ver-tical profile.

(iv) The different peaks assumed in the vertical pro-file can also be resolved by using the above al-gorithm.

(v) The above computations showed that by in-creasing the number of wavelengths and num-ber of observations it is possible to increase theresolution of the vertical profile.

(vi) The ozone vertical profile derived with the al-gorithm explained in this paper using visible re-gion of solar spectra gives ± 15% error com-pared to the Umkehr derived ozone profile.

t·'.

AcknowledgementThe authors wish to express their sincere thanks

to Prof. D R Sikka and Dr A S R Murty for givingvaluable suggestions and encouragement during theperiod of this study.

References1 Bose S, Trimbake H K, Londhe A L & Jadhav D B, High Res-

olution UV- Visible Spectrometer for Atmospheric Studies inSci Rep, No. RR-045, Indian Institute of Tropical Meteoro-logy, Pune, 1991,pp. 1-28.

2 Solomon S, Schmeltekopf A L & Sanders R W, J Geophys Res(USA), 92 (1987) 8311.

3 Sanders R W, Solomon S, Carroll M A & Schmeltekopf A L, JGeophys Res ( USA), 94 (1989) 11381.

4 Solomon S, Mount G H, Sanders R W & Schmeltekopf A L. JGeophys Res (USA), 92 (i987) 8329.

5 Noxon J F, Whipple E C (Jr) & Hyde R S, J Geophys Res(USA), 84 (1979) 5047.

6 DeMore W B, Margitan J J, Molina M J, Watsan R J, GoldenD M, Hampson R F, Kurylo M J, Howard C J & RavishankaraA R, Chemical Kinetics and Photochemical Data for Use inStratospheric Modelling, Evaluation No.7, JPL Publication,pp. 85-37, 1987.