13. SOLAR CONSTANCY13.1 Introduction
The variation of ozone amount in response to solar activity and solar variations on various time scales is of great intrinsic interest, and as well, provides valuable tests of atmospheric chemistry models, especially their radiation components. It is important therefore to consider to what extent the ozone measurement method itself is affected by, and is in error owing to, these solar variations. This discussion concerns the Dobson instrument only, but the problem also affects satellite ozone instruments which use extraterrestrial intensities or intensity ratios which are fixed between periodic recalibrations.
The accuracy of Dobson instrument ozone estimates depends on the constancy of solar extraterrestrial irradiance through the expression:
d(Lok) dXk = -------- (13.1) μhΔαkwhich is derived from equation (1.8). Constancy in absolute irradiance is of no significance of course. Constancy is demanded only for the irradiance ratios of bandpairs, or for the ratio of irradiance ratios when a double bandpair method, such as the AD method, is used. Note that, once again, the double bandpair practice tends to reduce the susceptibility to error. The ozone uncertainty described by equation (13.1) depends inversely on the airmass and the relative absorption coefficient, whose product in normal practice ranges from about 1.5 to 4.5 atm-cm-1. All instruments are affected in the same way, but there will be a tendency to find larger values of μhΔαk at higher latitude sites, and hence smaller effects there.
If the ozone uncertainty is to be kept to 0.0015 atm cm, i.e., about 0.5%, then the uncertainty in the intensity ratio must be less than about 0.5% also. The constancy of 0.5% is required only over the interval between the periodic recalibration of the instrument's extraterrestrial constants. These intervals are typically two to five years, but may be much longer. Constancy is always assumed, but the evidence to support the assumption is rather limited.
13.2 Solar spectral irradiance variations
The visible and near UV parts of the solar spectrum are emitted by the sun's photosphere, the apparent "surface" of the sun. The absorption of upwelling radiation within the photosphere by overlying gas restricts the perceived emitting region to a relatively shallow layer. The radiation emitted toward Earth from the outer parts of the solar disc is subjected to more absorption along its slanted photospheric path and therefore comes from higher and cooler layers and is less bright. This relative darkening near the disc periphery is known as limb darkening. It is spectrally dependent and is most marked in the near UV. Subsequent absorption by highly ionised material in the sun's chromosphere imposes strong and spectrally dense absorption lines, known as Fraunhofer lines, on the near UV spectrum (see Figure 5.1). Overall, the near UV spectrum incident at Earth is roughly similar to that of a 5700°K blackbody.
The sun rotates with a period which varies from about 27 days at its equator to about 24 days at its poles. Large darker and cooler areas of the sun's surface, known as sunspots, appear from time to time and rotate across the disc. These are magneto-hydrodynamically active zones whose numbers and size vary in a cyclical fashion with a period of about 11 years. The maximum of the last cycle occurred about 1979.
Considerable efforts have been made, especially in recent decades, to measure the solar spectrum and its variations. (For an extensive summary of the field, see White, 1977). In the UV, visible, and near IR regions, these efforts have concentrated mainly on the following:
firstly, the measurement of the solar constant (the total irradiance at mean Earth distance) to high accuracy (better than 1%);secondly, the measurement of the spectral distribution of this energy to moderate accuracy (about 2 to 5% in the near UV);
thirdly, the measurement of the middle UV spectral fluxes which are of importance to aeronomy, to accuracies of 10% to 50%.
Thus, most of the available spectral data can only resolve variations greater than 5%, which is not sufficient for our problem. There are many experimental difficulties. The long term stability of calibrations is always difficult to maintain. Measurements made on the centre of the solar disc and corrected for limb darkening may differ from full-disc measurements. The Fraunhofer line structure in the 300 to 350 nm region can result in intensity gradients of up to 40% nm-1 for 1 nm spectrometer bands, which means that wavelength accuracy is very important, and it can result in significant measurement differences between instruments of different bandwidths. Despite these limitations, a brief summary of the relevant available data is attempted as follows.
A variety of measurements of the solar constant made since 1976 agree to within 0.3% of their mean result (Frohlich, 1982). Individual measurement methods indicate small trends, of the order of 0.03% yr-1, and satellite measurements show continual variations on the time scale of days and weeks, mainly dips of the order of a few tenths of percent. These appear to be due largely to sunspots, which are considered to block part of the outgoing flux. The energy thus not radiated appears to be stored in a very large reservoir of very long time constant, and hence to contribute to very small long-term variations (Frohlich, 1982). The wavelength dependence of these variations is unknown. Some wavelength dependence of such sunspot-related variations should be expected owing to the wavelength dependences of limb darkening and of the emission with solar altitude.
Livingston's (1978) measurements of temperature-sensitive Fraunhofer lines at 538 nm imply a 6°K decrease in photospheric temperature in 1977, in parallel with the rise in sunspot number that year. However, Heath (1980) states that Livingston's subsequent measurements showed that this decrease continued right through the sunspot maximum, and that there were corresponding increases for other upper photospheric layers. It can be noted that any overall photospheric temperature increase of greater than 6°K would produce total flux changes which would be inconsistent with the known stability of the solar constant since 1976.
Heath and Thekaekara's (1977) summary of solar ultraviolet measurements showed a variation apparently related to the 11 year sunspot cycle, with amplitudes decreasing smoothly from about 50% at 160 nm to 10% at 300 nm. However, Simon (1978) considers that experimental uncertainties preclude the drawing of such a conclusion, and Nicolet (1980) states that there is no astrophysical basis for any significant variation above 175 nm. Moreover, Cooper's (1982) analysis of the semi-diurnal (12 hr) Earth atmospheric tide indicates that the UV band absorbed by ozone (approximately 200 to 300 nm), which provides the tide's forcing, varies by less than 2% over the sunspot cycle. Simon et al. (1982) conclude that the irradiance at 300 nm is known to ±5%. Heath and Thekaekara (1977) show that the irradiance variations associated with the 27 day solar rotation are less than 1% at 300 nm.
A more recent review of UV flux variability (Heath, 1980), especially as measured by satellite instruments, indicates a number of systematic variations in spectral irradiance. Among other things described are: firstly, differences between active and inactive parts of the 27 day cycle amounting to several percent below 285 nm but less than 0.5% above 300 nm; secondly, a variation over a fourteen month period which amounted to about 8% at 200 nm, 3% at 300 nm and less than 0.5% at 350 nm, and which, if real, would contribute a 0.06% yr-1 decrease to the solar constant (the possibility of instrument spectral drift was noted, however), thirdly, a long-term 15% amplitude variation in the 340 to 380 nm flux, relative to the 390 nm flux, deduced from ground-based Langley method measurements in 1931 and 1951 and satellite measurements in 1978. Frohlich and Wehrli (1982) give evidence of UV irradiance variations of the order of hundredths of a percent over periods of minutes and hours, which indicate that the amplitudes of the variations at 368, 500 and 778 nm are in the ratios of 4:2:1 approximately.
Some upper limits to the variation of the solar UV irradiance ratios may be determined from the measurements of the Dobson instrument itself. On very clear days with settled synoptic conditions, AD ozone estimates may be made which are repeatable to 0.5%, which suggests that variations in the irradiance ratio for the AD band combination are less than about 0.5% over the periods of tens of minutes for which repeatability holds good. Over longer periods it is necessary to exclude the effects of the atmosphere by studying the experimentally determined extraterrestrial constants as a function of time. The instrument component of the constants is checked by reference to standard lamps: thus the constancy of the solar intensity ratio is calibrated, in effect, against the constancy of the lamps' intensity ratio. Day to day variations in extraterrestrial constants during a period of their determination at Mauna Loa (Dobson and Normand, 1962) showed the solar irradiance ratios were constant over this period to at least 1%. It is possible that some institutions have maintained standard lamps with documented histories for many years and that from these the long term constancy of their Dobson extra-terrestrial intensity ratios could be determined. Such studies would be very valuable.
Incidentally, the long-term behaviour of ozone data can be used to set upper limits to the longer term variations of the irradiance ratios, though of course these limits are of no use in assessing the long-term accuracy of the ozone data. For example, yearly average ozone amounts for the world generally lie within 2% of the long term mean (WMO, 1981), which suggests that year to year variations in the irradiance ratios are less than approximately 4%. This estimate is complicated by the direct effect of any UV flux changes on ozone amounts, however.
The discussion of near UV solar flux variations given above is neither extensive nor conclusive. The independent evidence from balloon, rocket and satellite instruments does suggest that there are small variations on a variety of time scales, but generally the available data have large uncertainties, are uncorroborated, and are lacking in firm astrophysical explanation. Furthermore, the measurements available do not have the combination of spectral resolution and irradiance accuracy required to meet our specific needs. Dobson instrument measurements apply to the right bands and have sufficient precision, and do provide some estimates of the limits to the variations, but they are limited by uncertainties in instrument calibrations and atmospheric attenuation. Further advances with the problem must await the results of currently deployed or planned satellite instrumentation.
Overall, the evidence points to variations in the solar irradiance ratios for the Dobson bandpairs of less than 1%, and hence to resulting variations (errors) in the ozone estimates of less than 1%. However, the possibility of much larger variations must be acknowledged.
13.3 Summary
(i) The Dobson measurement method assumes that the solar irradiance ratios for its bands are constant (to less than about 0.5%) for the duration of the interval between recalibrations of extraterrestrial constants. The standard AD double bandpair method is insensitive to any irradiance change which is spectrally linear.(ii) Solar radiation in the 300 to 350 nm region is of photospheric origin, but is overlaid with strong chromospheric Fraunhofer absorption lines. The solar constant, which is also largely photospheric in origin, is known to 0.3%, but varies by a few tenths of a percent over daily and weekly time scales (probably in response to sunspot variations), and may have a long term trend of a few hundredths of a percent per year.
(iii) Measurements from satellites, rockets, balloons and ground-based spectrophotometry indicate that any near UV flux variations are of the order of a few percent or less. These data do not have the spectral resolution or accuracy needed for this study of the Dobson instrument.
(iv) Dobson instrument measurements indicate that irradiance ratios are constant to 1% or better over periods of minutes to days, and are probably constant to at least 4% on a year to year basis. Further study of Dobson instrument data and calibration records to assess solar variability would be very desirable.
(v) The evidence available at present has many uncertainties, but overall, it indicates that errors in ozone estimates due to UV flux variability are most unlikely to be greater than 1%.
Return to Table of Contents
Forward to Conclusion