9.       OZONE ABSORPTION COEFFICIENTS

9.1      Introduction

          The accuracy required of the ozone absorption coefficients depends on the application. Within the Dobson network there is a need for consistency between the various measurement types and hence between the coefficients of the various bandpairs. Within a larger integrated observing system comprising many different sorts of instruments, both ground based and satellite based, there is a more general need for consistency of coefficients throughout the whole spectrum. Absolute accuracy is usually of little significance, but it can be important for some purposes, such as atmospheric chemistry and radiative transfer studies. The temperature dependences of the coefficients must also be considered.

          Several papers in 1980 focussed on the question of the coefficients' accuracy. These papers, by Klenk (1980), Komhyr (1980a) and DeLuisi (1980), and a brief review of them by Mateer (1981), showed significant inconsistencies between the independent laboratory absorption data, and between the coefficients used by different instrument types and by different bands within individual instrument types. Also, in Komhyr (1980a) it is argued that the standard Dobson AD ozone estimation is too large by about 5% and that this is probably due to error in the standard AD coefficient. Before considering these papers further, it is necessary to review the considerable amount of work done on the topic before 1980.

9.2      Laboratory spectral absorption measurements

          The ultimate source of ozone absorption coefficients is the laboratory measurement of the spectral transmission of a chemically determined quantity of ozone. An absorption coefficient spectrum is shown in Figure 1.1. The measurements are by no means simple. The concentration, temperature and purity of a range of ozone samples must be accurately determined, the linearity and stability of the detection method must be ensured, the monochromator's wavelength calibration and stability must be verified, and stray light in the equipment must be rigorously rejected if it is not to swamp the very small amount of energy in the very narrow bands (less than 0.1 nm) being measured. The situation is in some ways similar to that of the field measurement of atmospheric ozone by UV optical means. However it would appear that the strenuous efforts made over many years to investigate and improve the field instruments, Dobson and other, have not been sufficiently well supported by similar efforts in the laboratory.

          The most widely used ozone absorption coefficient data are those of Vigroux (1953). We should not be too surprised to find that his data have been found wanting when applied to the very specific purpose of atmospheric ozone measurement with the Dobson instrument, since the scope of his experiments was very broad. The main aims were to investigate the wide differences in absorption coefficients measured by earlier workers, notably Buisson and Fabry (1913) and Ny and Choong (1933), to create an homogeneous set of coefficients over a wide and well-detailed spectral range from 230 nm to 10,000 nm, and to investigate the temperature and pressure dependences of the coefficients. Indeed Vigroux (1967) later clearly stated that the measurements were made without the least knowledge of their possible application to the Dobson instrument. Ozone absorption coefficients derived from the Vigroux (1953) data at -44°C, corresponding to stratospheric temperatures, were used for the Dobson instrument from 1 July 1957 to 31 December 1967. They were found to be about 36% smaller than the coefficients previously in use which were derived from Ny and Choong (1933) (Dobson, 1957).

          Also in 1953, measurements were made in the UV and visible regions by Inn and Tanaka (1953), but only at 18°C. Their results at this temperature differ from Vigroux's by up to 7% in some wavelength ranges, and no use has been made of them in the Dobson instrument. Further laboratory measurements have since been made by Hearn (1961), DeMore and Raper (1964), Vigroux (1967), Griggs (1968) and Simons et al. (1973), and another effort currently underway has been reported by Bass and Paur (1981). Vigroux's 1967 paper was specifically aimed at the Dobson instrument and shall be discussed in following paragraphs. Limited comparisons and discussions of these data sets may be found in Hudson (1971 and 1974) and Klenk (1980). Klenk's tabulations of band-integrated coefficients, some of which are reproduced here in Table 9.1, show differences of up to 15% for Dobson bands in the 300 nm to 340 nm region. It should be noted that all of the available data have limitations of one sort or another, particularly those of limited spectral range, limited spectral density of measurements, and limited temperature range, which restrict their accuracy or their direct application to the Dobson instrument bands. E. Vigroux (personal communication) has pointed out that the spectral density of his 1953 data generally did not justify the precision implied by the standard 1957 to 1967 coefficients. He states that for the short wavelength C band, values of 0.8 or 0.9 atm-cm-1, depending on the interpolation used, would have been as equally valid as the standard 0.865 atm-cm-1 value.

TABLE 9.1 Comparison of Dobson instrument ozone absorption coefficients and their combinations. All data are corrected to either -44°C or -50°C.
Type Laboratory Spectra Dobson experiments Ad hoc Units
Original Source
Original Temp
Obtained from
VIG53
-44
DOB57
VIG67
18
VIG67
VIG67
18
BSH
ITN53
18
KLK80
SIM73
18
KLK80
DOB63
18
lab
VIG67
18
lab
W&P71
-50
lab
DOB63
atmos
Oxf'd
DOB63
atmos
Edm'n
IOC68
 
W&P71
KMH80
 
KMH80
°C
Wavelength, nm
305.5 
308.8 
311.45
317.6 
325.4 
329.1 
332.4 
339.8 
 
1.882
1.287
0.912
0.391
0.120
0.064
0.047
0.017
 
1.863
1.276
0.850
0.377
0.117
0.070
0.043
0.012
 
1.871
1.284
0.858
0.373
0.112
0.064
0.044
0.012
 
1.999
1.330
0.922
0.400
0.123
0.075
0.050
0.017
 
2.139
1.400
1.014
0.422
0.123
0.087
0.050
0.021
 
-
-
-
-
-
-
-
-
 
-
-
-
-
-
-
-
-
 
-
-
-
-
-
-
-
-
 
-
-
-
-
-
-
-
-
 
-
-
-
-
-
-
-
-
 
-
-
-
-
-
-
-
-
 
1.976
1.287
0.903
0.395
0.120
0.064
0.047
0.017
 
atm-cm-1
 "   " 
 "   " 
 "   " 
 "   " 
 "   " 
 "   " 
 "   " 
A  
B  
C  
D  
AD 
CD 
1.762
1.223
0.865
0.374
1.388
0.491
1.746
1.206
0.807
0.365
1.381
0.442
1.759
1.219
0.814
0.360
1.398
0.454
1.876
1.255
0.872
0.383
1.493
0.489
2.016
1.313
0.964
0.401
1.615
0.563
1.742
1.142
0.808
0.354
1.388
0.454
1.756
1.133
0.799
0.360
1.396
0.439
1.748
1.159
0.810
0.360
1.388
0.449
1.741
1.144
0.791
0.353
1.388
0.438
1.742
1.155
0.804
0.354
1.388
0.450
1.748
1.140
0.800
0.360
1.388
0.440
1.856
1.223
0.856
0.378
1.478
0.478
atm-cm-1
 "   " 
 "   " 
 "   " 
 "   " 
 "   " 
A/D
B/D
C/D
4.711
3.270
2.213
4.784
3.304
2.211
4.886
3.386
2.261
4.898
3.277
2.277
5.027
3.274
2.404
4.921
3.226
2.282
4.878
3.147
2.219
4.856
3.219
2.250
4.932
3.241
2.241
4.921
3.263
2.270
4.856
3.167
2.222
4.910
3.235
2.265
 
Note: The author abreviations are: VIG-Vigroux, ITN-Inn and Tanaka, SIM-Simons et al., DOB-Dobson, W&P-Walshaw et al., and Powell, KMH-Komhyr, KLK-Klenk, and BSH-Basher. All the sources are discussed in the text. The purpose of including the author's (BSH) calculations is to show the degree to which interpretations of the Vigroux 1967 data may differ. Those sets scaled to give the AD coefficient equal to 1.388 are shown by underlining thus 1.388.

9.3      Investigations using the Dobson instrument

          Indirect information on the accuracy of the absorption coefficients used with the Dobson instrument may be found from the consistency of the ozone measurements made with different pairs of bands. A preliminary discussion of this approach is made in Dobson and Normand (1962). A basic limitation is the lack of information on the exact effect of aerosol spectral attenuation on the derived coefficients (Basher, 1976). Dobson (1963) reported 10% differences between different band pairs and used data from days with clear skies at Oxford, England to estimate a consistent set of coefficients, scaled such that the AD band combination's coefficient remained as before at 1.388 atm-cm-1. These coefficients differed from the standard coefficients by -1.2%, -6.5%, -8.6% and -5.6% for the A, B, C and D bandpairs respectively. This approach was followed by a number of others whose results, summarised by Shah (1968), show differences, one with another, of several percent, this probably being due to aerosol scattering effects and to differences in instrument adjustment and calibration.

          An alternative method of determining the absorption coefficients appropriate to the Dobson instrument bandpairs is to use the instrument itself to measure the coefficients directly in the laboratory. Measurements of this sort by Dobson (1963) at 5°C and 50°C, when corrected to -44°C and scaled such that the AD band combination's coefficient was 1.388 atm-cm-1, agreed to within 2% of his field measurements referred to above. No measurement of laboratory ozone amount was made. Vigroux (1967) carried out a similar experiment at Mont-Louis in the French Pyrenees at about 18°C and using direct sunlight as a source. The absolute amount of ozone was determined from near-simultaneous photographically recorded transmission spectra together with the spectral absorption coefficients for 18°C found by Vigroux (1953). The correction of the coefficients to the -50°C temperature chosen was made using the temperature dependences also reported by Vigroux (1953). The resulting ozone absorption coefficients were found to differ from the standard coefficients by 0.6% for the AD band combination and by -0.3%, -7.4%, -7.6% and -3.7% for the A, B, C, and D bandpairs respectively. Notice that these differences are very similar to those found by Dobson (1963). Vigroux also redetermined the actual absorption coefficient spectra in much finer detail in the vicinity of the Dobson bands using the photographic spectrograph and again using the earlier 1953 work for temperature correction and absolute calibrations. The differences between the coefficients calculated from these "fine structure" measurements and the standard coefficients were -0.5% for the AD band combination coefficient and -0.9%, -1.4%, -6.7%, -2.4% for the A, B, C and D bandpairs respectively. The result for the B bandpair is not consistent with the direct measurements of Vigroux or Dobson, but the A, C and D results agree satisfactorily.

          The evidence available in 1967 was sufficiently compelling that the International Ozone Commission recommended a new set of coefficients, 1.748, 1.140, 0.800 and 0.360 atm-cm-1 for the A, B, C and D bandpairs respectively to be used from 1 January 1968 onward. They are still currently used. The 1.388 atm-cm-1 value for the AD bandpair combination was retained partly to ensure the continuity of the data records, and partly because the new Vigroux (1967) values of 1.396 atm-cm-1 and 1.381 atm-cm-1 were very close to the existing value. The evidence does not support Komhyr's (1980a) view that the retention of 1.388 was "highly arbitrary." At the same time, it can be noted that this standard of absolute calibration was totally reliant on the Vigroux (1953) measurements which show some differences with other experimenters' data sets, and that there remained an uncertainty of about 5% in the Vigroux (1967) coefficients for the B wavelength pair, which, even today, remains unexplained (Vigroux, personal communication).

          Further efforts to determine mutually consistent ozone absorption coefficients using the Dobson instrument in the laboratory at low temperatures were made by Walshaw et al. (1971) and Powell (1971). The results show excellent agreement one with another, and their mean values estimated by the author for -50°C differ from the standard 1968 values by 1.5% or less. Walshaw et al. also raise the question as to the importance of the light source used for determining effective ozone absorption coefficients with the Dobson instrument. They state that Vigroux considered it important to use direct sunlight since the weighting of the coefficient spectrum by the spectrally very variable solar spectrum would give different results to those found when an incandescent lamp is used. Measurements by Dobson (1963), Walshaw et al. (1971) and Powell (1971) on lamps and sunlight showed no differences, however. The author has investigated this question by calculating the band-integrated coefficients that would be measured for various sources, using the 0.05 nm spectra described in Section 5.2. Table 9.2 lists the coefficients for a 3000°K blackbody lamp and the deviations from these values for direct sunlight at various conditions of airmass and ozone. The results give support to both sides of the arguments on the one hand relatively large deviations, of up to 6%, do exist for some wavelength bands, but on the other hand, the deviations for the standard bandpairs and particularly the AD band combination are relatively small.

TABLE 9.2

Calculated effect of the light source used in the determination
of Dobson ozone absorption coefficients.

Wavelength or bandpair Lamp, 3000 K blackbody. Coefficient and its percent difference from the unweighted value. Direct sunlight, for airmass mh and total ozone X. Percent difference relative to lamp coefficients
μh = 0
 X = 0
μh = 1
 X = .2
μh = 1
 X = .3
μh = 1
 X = .4
μh = 2
 X = .3
μh = 3
 X = .3
nm atm-cm-1 % % % % % % %
 305.5
 308.8
 311.45
 317.6
 325.4
 329.1
 332.4
 339.8
1.8644
1.2823
0.8549
0.3692
0.1141
0.0648
0.0425
0.0124
-0.04
-0.06
-0.08
-0.09
-0.50
-0.81
-0.66
 0.00
 0.20
 0.21
 0.56
 1.44
-1.14
-2.35
 1.69
 0.11
 0.11
 0.02
 0.33
 1.22
-1.87
-3.35
 1.06
-0.21
 0.08
-0.06
 0.22
 1.12
-2.14
-3.70
 0.87
-0.26
 0.04
-0.15
 0.12
 1.02
-2.41
-4.04
 0.68
-0.31
-0.04
-0.36
-0.13
 0.78
-3.14
-5.00
 0.06
-0.42
-0.15
-0.67
-0.49
 0.44
-4.14
-6.25
-0.73
-0.57
 A
 B
 C
 D
 AD
1.7503
1.2175
0.8124
0.3568
1.3936
-0.01
-0.02
-0.05
-0.10
 0.02
 0.29
 0.35
 0.50
 1.50
-0.02
 0.24
 0.20
 0.29
 1.27
-0.02
 0.22
 0.13
 0.19
 1.17
-0.02
 0.20
 0.06
 0.09
 1.06
-0.02
 0.16
-0.11
-0.14
 0.82
-0.01
 0.11
-0.38
-0.48
 0.47
-0.02

9.4      Temperature dependence of absorption coefficients for the Dobson instrument

          Ozone's absorption in the 300 nm to 340 nm wavelength region increases with temperature (Vigroux, 1953; Simons et al., 1973). Vigroux's measurements, made at nine temperatures, show that the principal minima increase steadily with temperature, by as much as 0.7%°C-1 for deep minima beyond 315 nm, whereas the principal maxima show a turning pattern, at first decreasing, by about 0.1%°C-1 from -92°C to -75°C, remaining relatively constant from there until -44°C, then increasing, slowly at first, then more rapidly. Fortunately, the dependences are relatively low for the important region of 305 nm to 318 nm and for stratospheric temperatures. Band integrated dependences have been estimated for the A, B, C and D bandpairs respectively at 0.11, 0.10, 0.18 and 0.12%°C-1 by Walshaw et al. (1971) and at 0.14, 0.17, 0.18 and 0.27%°C-1 by Thomas and Holland (1977). The differences are indicative of the uncertainty of the estimations. Walshaw et al. gave histograms of temperature at the region of the main ozone maximum for 1965 at four sounding stations, while Thomas and Holland integrated mean ozone temperature from a number of standard ozone and temperature profiles, and both arrived at a maximum range of about -40°C to -65°C. For this range the errors in mean absorption coefficient would be about ±1% to ±3%, depending on the bandpair, and about ±l.5% for the AD band combination. Any changes in the mean profiles of both temperature and ozone induced by changing amounts of chlorofluorocarbons and carbon dioxide might conceivably result in a systematic change of 5°C in the mean ozone temperature and therefore of 0.5% in the mean AD absorption coefficient. It can be noted that at present there is no firmly established reference temperature for which coefficients are defined and there are few, if any, experimenters who correct for changing mean stratospheric temperature.

9.5      Absolute accuracy of Dobson absorption coefficients

          The strict definition of band-integrated absorption coefficients (equation 6.1) involves weightings of the laboratory absorption data by the spectra of solar irradiance, atmospheric transmittance and instrument transmittance. However the Dobson instrument's fixed coefficients are usually calculated from just simple weightings by nominal triangular or trapezoidal slit transmittance functions. This leads to small errors, since solar irradiance and atmospheric transmittance are spectrally very variable, and the true shape of the Dobson transmittance functions are poorly known (see Sections 2, 5, and 6). Further errors arise from the interpolation of the currently available absorption spectra in data-sparse regions, and from the bandwidth effect (see Section 6). Overall, these various factors may contribute errors of up to 1 to 3%, depending on the circumstances, which are in addition to the errors associated with the laboratory data and their temperature dependences. The differences between the coefficients calculated by Vigroux (1967) and by the author, and shown in Table 9.1, under VIG67 and BSH respectively, give some indication of the size of the errors.

          Klenk's (1980) principal aim was to determine the most suitable ozone absorption coefficients for use with the satellite back-scattered ultraviolet (BUV) experiments. He took each of the laboratory data sets of Vigroux (1953 and 1967), Inn and Tanaka (1953) and Simons et al. (1973), and computed band averaged coefficients for both BUV and Dobson instruments at -44°C, correcting the latter two sets from 18°C and -78°C respectively, with the aid of the temperature dependences determined by Vigroux (1953). The results are reproduced in Table 9.1, along with other comparative data. Differences between the Vigroux data and the Inn and Tanaka data are generally less than 6%, but differences between these data and those of Simons et al. are larger, rising to 25% in one case. The only independent test of the merit of the various coefficients is that of consistency, as evidenced for example by the A/D, B/D, and C/D ratios in the table, and by the data comparisons given by Klenk of Dobson CD versus AD band combinations, BUV-B bandpair versus Dobson AD, and BUV-A versus BUV-B bandpairs. These tests indicate that the Inn and Tanaka coefficients (for -44°C) are at least as consistent as the Vigroux (1967) coefficients. All of the band-integrated data will suffer from errors of interpolation of course.

          In his examination of the absolute accuracy of the Dobson instrument, Komhyr (1980a) compared sets of ozone measurements made by the Dobson instrument with those made by other types of ozone measuring instruments, and concluded on the basis of five separate comparisons that the Dobson AD measurements may be too high, by about 5%. However, it has since been shown that each of the comparisons is subject to large uncertainties, and that for some, the data were not in fact independent (Basher, 1982b). There still remains the possibility of a small positive bias, though with an uncertainty of several percent. The preliminary NBS absorption cross sections now also indicate a small positive bias (A.M. Bass, W.D. Komhyr, personal communications).

          The differences between the 1957 and 1968 standard A, B, C. and D absorption coefficients, and therefore between their resulting ozone estimates, cover a range of about 7%, while the corresponding differences for the paired bandpairs, e.g. AD, BC, etc., extend to a range of 23% Komhyr (1980a). No special explanation is required for the large 23% range, since, as is discussed by Basher (1982b), its larger size is due solely to the linear combination of the differences of from +1% to +8% found in the individual A, B, C and D bandpair coefficients. These +1% to +8% differences are comparable with, and are almost certainly mostly due to, the uncertainties in the laboratory absorption data.

          DeLuisi (1980) used the measurements of a UV double monochromator to estimate ozone amounts as a function of wavelength. The results, displayed as a spectrum of departures from the mean of the values between 298 to 313 nm, show a definite structure with an RMS departure of about 0.010 atm cm in ozone, or about 3%. This is twice the estimated measurement precision. The structure may be partly due to the bandwidth effect and the sensitivity of the ozone absorption coefficients and extraterrestrial irradiance to uncertainty in the instrument's wavelength calibration, since these are very spectrally dependent and could easily contribute spectrally varying errors of a few percent (Basher, 1982b). However, as suggested by DeLuisi, the error in the Vigroux (1953) absorption coefficient data used to derive the ozone estimates must be considered a primary cause.

          Calculations of band-integrated coefficients by the author (see Section 5) for the Brewer grating instrument show agreement to better than 2% with the atmospheric estimations and the laboratory determinations of Kerr et al. (1977) for their band combinations "2,4" and "2,3,4", but for those band combinations involving the important shortest wavelength number 1 band, at 309.95 nm band, the calculated values were all too high within the range of 0.096 ±.004 atm-cm-1, or about 10% in the coefficient at 309.95 nm. This suggests a systematic error at about 309 nm, either in the original Vigroux data, or in their interpolation by the author. Shaw (1979) found that his filter photometer measurements in the Chappuis band (500 to 700 nm) at Mauna Loa gave ozone amounts 17% lower than concurrent Dobson AD ozone measurements, though this may simply reflect the unsuitability of the Chappuis band for ozone measurement, or an even lower accuracy of the absorption coefficients in that band.

          It is clear that few improvements in the absolute accuracy of ozone measurements can be expected without further improvement in the available laboratory ozone absorption coefficient data. Fortunately, the ozone absorption measurements by Bass and Paur (1981) at the United States National Bureau of Standards (NBS) are being made for just this purpose, and are now becoming available. The International Ozone Commission has recently established (in 1982) a working group to thoroughly examine the absorption coefficient question, and if possible, to recommend new Dobson coefficients based on the NBS data (C.L. Mateer, personal communication). The working group will, among other things, assess the accuracy of the NBS data and directly measure the spectral transmittances of Dobson instrument bands. McPeters and Bass (1982) report an accuracy of 2% or less for the NBS data. This, together with the various other uncertainties in the calculation of band?integrated coefficients discussed in this Section, and in Basher (1982b), suggests that the total uncertainty in the effective coefficients of UV ozone instruments may become about 2% in the near future.

9.6 Summary

(i)      Published ozone absorption coefficient spectra mesured by independent investigators can differ by 10% or more in the 300 nm to 340 nm wavelength region. The implied magnitude of systematic error in the determination of atmospheric ozone is intolerable for present-day measurement needs.

(ii)     The self-consistency of the Dobson instrument's standard A, B, C, and D bandpair absorption coefficients appears to be better than 2%.

(iii)    There is some evidence that the absolute values of the standard absorption coefficients used for the Dobson instrument may be too low by a few percent, i.e., that Dobson ozone column amount estimates may be too high by a few percent.

(iv)     The accuracy of band-integrated ozone absorption coefficients is limited by their dependences on the method of interpolating laboratory absorption data, on stratospheric temperature, and on the detailed UV spectral characteristics of the sun, the atmosphere and the instrument. The dependence on mean ozone layer temperature limits the accuracy of the fixed AD coefficient to about ±l.5%.

(v)      The agreement between spectrally independent instruments in the UV region (e.g., BUV instrument versus Dobson instrument) at present cannot be expected to be better than about 3%.

(vi)     A greatly improved accuracy, of 2% or less, has been reported for the new laboratory absorption measurements of Bass and Paur (1981). These data are being examined by the International Ozone Commission with a view to recommending a new, more accurate set of coefficients for the Dobson instrument.


Return to Table of Contents

Forward to Atmospheric Scattering