7. FIELD OPERATIONS AND CALIBRATIONS7.1 Introduction
The purpose of this section is to discuss those aspects of field operations which affect the accuracy of the Dobson network's observations. Particular attention is given to organisational factors, which are of importance owing to the international nature of the network and the manual character of the instrument, and to the determination of extraterrestrial constants, which is well known to be an important source of error.
7.2 Organisational factors
The present global Dobson ozone monitoring network has evolved slowly over many decades, and it continues to be dependent on the cooperation and goodwill of many scientists and many countries. The approximately sixty instruments in use throughout the world are owned and operated by a variety of institutions, often national meteorological services or associated research institutes, and so do not constitute an homogeneous, centrally organised network. Some are operated without much support or encouragement from the institution concerned, which may see little benefit in such long term specialist monitoring to its own current research work or other goals. Clearly there is the potential for wide variation of operational practices.
The instrument is manually operated. This means that there is a decidedly human element to its operation, and that considerable attention must be given to personnel matters to ensure a successful operation. Staff must be trained in routine operations, and for some, in calibration work. The instrument and the procedures used must be inspected periodically. The procedures themselves must be well thought out and clearly described. There should be rapid feedback to the operator when quality control measures show that the performance of either the instrument or the operator is not up to standard.
It is important that senior officers and scientists take an active interest in the observers' work and convey to them something of the international significance of their observations. The Dobson observations are likely to comprise only a small part of the observers' duties, and may be accorded a lower priority if other immediately pressing duties or demands arise. Also, the level of experimental care and judgement required may be inconsistent with the other duties. To compound the difficulties, in some organisations the staff may change frequently according to shift schedules and as promotions and other movements occur.
Before 1957, instructions for calibrating and operating the instrument were provided by the manufacturer using material prepared by G.M.B. Dobson. In that year, more detailed sets of instructions (Dobson, 1957a and 1957b) were published, as part of an International Geophysical Year instruction manual. These were aimed at fairly expert readers, and it is probable that, in due course, many Dobson instrument scientists prepared simpler, more fully explained instructions for the use of their operators. For example, a comprehensive instruction manual (Komhyr, 1962) was prepared for use in the United States network. More recently, the WMO Operations Manual (Komhyr, 1980b) was prepared in order to encourage more modern and uniform standards of operation throughout the whole network. It is not clear to what extent the WMO manual's procedures are being followed at present.
The international intercomparisons of instruments, and the designation of the Central Dobson Spectrophotometer Laboratory at Boulder, Colorado, have done a great deal to improve the consistency of instrument calibrations. What is now required is a similar international effort to improve the consistency and quality of routine operations. The production of the WMO Operations Handbook was an important step in this direction. Other activities which could be considered are:
- a survey of current operational practices;Ideally, any such improvements would be best set in the context of a comprehensive quality assurance programme designed specifically for the network using the principles of quality assurance already established in the manufacturing industries.- the preparation and regular distribution to all stations of standard forms for such things as defining times of observations, for reporting results, etc.;
- the training of operators or their supervisors at the Central Laboratory or at regional centers;
- the further development and dissemination of improved operating practice;
- and possibly the circulation of an operations newsletter.
7.3 Field calibrations and tests
An important source of error, especially in past data, is that arising from the uncertainty in extraterrestrial constants determined independently at each site by the Langley method. The method is briefly described in Section 1, and further details are to be found in Sections 4 and 5. Before the mid-1970s, when direct intercomparison calibrations became established, all instruments were calibrated by the Langley method, and even now perhaps a third of the network relies on such calibrations.
Extraterrestrial constants were supplied with each instrument upon delivery. These were obtained by G.M.B. Dobson at Oxford by Langley method measurements and by direct intercomparison with his own instrument, No. 1 (Dobson, 1968). However Dobson and Normand (1962) showed that although the Langley method worked well for the instruments at Mauna Loa and at certain sites in Canada, it gave inconsistent results at Oxford, i.e. it failed to give the linearity of data expected from equation 1.3 or the constancy expected of the extraterrestrial constants. They attributed this to the atmospheric conditions at Oxford, though stray light also could have played a large part (see Section 4).
The standard errors in the daily extraterrestrial constants reported by Dobson and Normand (1962) for the Mauna Loa and Canadian instruments imply average ozone errors of less than 1%. The ozone errors are not constant of course, but vary with relative absorption coefficient and inversely with airmass (see equation 1.4), and hence vary with wavelength pair, latitude, season and time of day.
For determinations made under changeable atmospheric conditions, or with poorly adjusted instruments, or with instruments having a stray light problem, the average error in AD ozone measurements should be similar to those shown by Dobson and Normand (1962) for Oxford, namely about 5% or less. Worst case average errors of 10% or more cannot be ruled out. In some cases the manufacturer's constants obtained at Oxford were used for many years before any independent Langley determinations were made. The size of these errors are consistent with the differences of up to 10% found in the early intercomparisons, and with the range of about 10% found in a recent set of mean differences of co-located Dobson and BUV-TOMS satellite observations (WMO, 1982).
It is instructive to look in detail at the results of the international Dobson spectrophotometer intercomparisons at Belsk, Poland in 1974, and at Boulder, Colorado, U.S.A. in 1977. Table 7.1 lists the published ΔNk corrections for the Belsk intercomparison (Dziewulska-Losiowa and Walshaw, 1977, Table I) along with the resulting percentage AD ozone corrections calculated for an ozone amount of 0.300 atm cm. This 1977 publication corrected some of the original data in Dziewulska-Losiowa and Walshaw (1975) and gave the dates of the wedge calibration tables to which the corrections apply.
TABLE 7.1 Published results of the international intercomparison of Dobson spectrophotometers at Belsk, Poland in 1974. Correction are to be added to old values. XAD corrections are calculated for an ozone amount of 0.300 atm cm. Instrument Country Required N-Value corrections Resulting XAD Corrections in % ΔNA ΔNC ΔND ΔNAD ΔNCD μ=1 μ=2 μ=3 Mean 41 U.K. .002 .006 -.014 .016 .020 3.9 1.9 1.3 2.3 64 G.D.R. -.216 -.212 -.244 .028 .032 6.7 3.4 2.2 4.1 77 Canada -.017 -.003 -.006 -.023 -.009 -5.5 -2.8 -1.8 -3.4 83 U.S.A ..... Reference instrument .... ... ... ... ... 84 Poland .048 .014 -.011 .059 .025 14.2 7.1 4.7 8.7 96 Egypt -.036 -.026 -.049 .013 .023 3.1 1.6 1.0 1.9 101 Switzerland .001 -.007 -.017 .018 .010 4.3 2.2 1.4 2.6 108 U.S.S.R. -.015 -.012 -.040 .025 .028 6.0 3.0 2.0 3.7 110 Hungary .077 .139 .148 -.071 -.019 -17.1 -8.5 -5.7 -10.4 112 India -.030 -.025 -.006 -.024 -.019 -5.8 -2.9 -1.9 -3.5 Note: The XAD correction are accurate to no better than ±0.5%The listed mean AD ozone errors, which are the mean of the errors for airmasses 1, 2 and 3, lie in the 2% to 10% range of magnitudes and reflect the level of accuracy to be expected of independently calibrated, independently maintained instruments. Some very large differences for individual bandpairs are shown. These and various other unsatisfactory features not evident in the final AD ozone corrections listed in Table 7.1 are discussed by Dziewulska-Losiowa and Walshaw (1975), and in paragraphs below.
The published results of the 1977 Boulder intercomparison (Komhyr et al., 1981a) are reproduced in Table 7.2. Considerable attention was given at this intercomparison to properly adjusting and calibrating, and in some cases modernising, the instruments' optics and electronics. The final correction factors include the effect of these changes as well as the errors in the extraterrestrial constants. As noted by Komhyr et al. (1981a), the comparability of the instruments' AD ozone estimation, i.e., of their standard ozone measurement type, was better than at the Belsk intercomparison, with five of the eight instruments needing corrections of only about 1% or less.
TABLE 7.2 Published results of the international intercomparison of Dobson spectrophotometers at Boulder, Colorado, U.S.A. in 1977. Correction are to be added to old values. XAD corrections are calculated for an ozone amount of 0.300 atm cm. Instrument Country Required N-Value corrections Resulting XAD corrections in % ΔNA ΔNC ΔND ΔNAD ΔNCD μ=1 μ=2 μ=3 mean 41* U.K. .0378 .0335 .0308 .0070 .0027 1.7 0.8 0.6 1.0* 71 G.D.R. -.0488 -.0721 -.0920 .0432 .0199 10.4 5.2 3.5 6.3 77* Canada .0373 .0275 .0450 -.0077 -.0175 -1.8 -0.9 -0.6 -1.1* 83 U.S.A. ..... Reference instrument ....... ... ... ... ... 96* Egypt -.0871 -.0536 -.0691 -.0180 .0155 -4.3 -2.2 -1.4 -2.6* 105 Australia .0069 -.0130 -.0119 .0188 -.0011 4.5 2.3 1.5 2.8 108* U.S.S.R. .0662 .0423 .0713 .0051 -.0290 -1.2 -0.6 -0.4 -0.7* 112* India .0286 .0315 .0270 .0016 -.0045 0.4 0.2 0.1 0.2* 116 Japan .0025 .0029 .0069 -.0044 -.0040 -1.1 -0.5 -0.4 -0.6 Notes: (i) * denotes those instruments previously calibrated against the no.83 reference instrument at Belsk. (ii) The accuracy of the N corrections is less than the number of decimal places implies. The XAD corrections are accurate to no better than ±0.5%
TABLE 7.3 Corrections arising from the error in sign (see text), to be added to the N-values of the World Reference Spectrophotometer No. 83 between its calibration dates in 1972 and 1976, and to any instruments calibrated against No. 83 during this period. Approximate equivalent ozone corrections, also to be added, calculated for an ozone amount of 0.300 atm. cm. Band N-value Approximate ozone correction, % Combination correction μ=l μ=2 μ=3 mean A 0.028 5.3 2.7 1.8 3.3 C 0.023 9.6 4.8 3.2 5.9 D 0.031 ------ Not relevant ------ AD -0.003 -0.7 -0.4 -0.2 -0.4 CD -0.008 -6.1 -3.0 -2.0 -3.7 Table 7.2 does show some rather disquieting features, however. In particular, the ΔNk corrections for the individual bandpairs of the instruments, No.s 41, 77, 96, 108, and 112, which had been previously calibrated against the No. 83 reference at the 1974 Belsk intercomparison, are surprisingly large. A typical value, 0.0400, represents an error in an intensity ratio of about 10%, and is much larger than the precisions available from either the independent Langley method, or the routine standard lamp checks.
Fortunately, there is a simple explanation for at least some of these large values. After the completion of the Langley calibration of the World Reference Spectrophotometer No. 83 at Mauna Loa Observatory in 1972, the required corrections to the instrument's extraterrestrial constants were inadvertently made using the wrong sign (Peterson, 1978). The effect on the standard AD ozone estimation was very small, and the error remained undiscovered until the results from the next calibration at Mauna Loa in 1976 were processed. The N-value corrections needed to account for these errors of sign (Peterson, 1978) are given in Table 7.3.
It is important that the nature of these corrections and their application be clearly stated. The significant points are as follows
- The errors in the standard AD ozone estimations, due to the reference error, average to about 0.4%. Therefore the great bulk of ozone data is affected only to about 0.4%, which is relatively small.- The errors in the A, C and CD ozone estimations, due to the reference error, are about 3%, 6% and 4%, respectively, and hence are significant.
- The calibration of the reference Dobson spectrophotometer No. 83 was in error at the 1974 Belsk intercomparison, by the amounts listed in Table 7.3. The results of the intercomparison, published by Dziewulska-Losiowa and Walshaw (1975 and 1977) and also listed in Table 7.1, have not been corrected for these errors.
- The No. 83 reference Dobson spectrophotometer was correctly calibrated at the time of the 1977 Boulder intercomparison, and the final calibrations of the participating instruments are correct.
- The final corrections which resulted from the Boulder intercomparison, and which appear in Peterson (1978), Komhyr et al. (1981a), and in Table 7.2 above, include, for those instruments which also participated in the Belsk intercomparison, a component which compensates for the reference error inadvertently imposed at the Belsk intercomparison.
- The reference error was reported to the 1977 Boulder intercomparison participants, at the time of the intercomparison, and to the World Meteorological Organisation (W.D. Komhyr, personal communication).
- The changes in the calibration of the No. 83 reference instrument arising from the 1972 Mauna Loa calibration are exactly half the corrections given in Table 7.2, and hence amount to about 0.2% for the AD ozone estimate and to a few percent for the other main band combinations.
Approximate mean ozone corrections for the Boulder intercomparison are shown in Table 7.4; (a) for the data as published, and (b) with the reference error component removed from those instruments which had also been intercompared at Belsk. As before, the corrections are calculated as the mean of the corrections for airmasses 1, 2 and 3, and for an ozone amount of 0.300 atm cm. The values remaining after the removal of the reference error component represent the intrinsic changes to the instruments' calibrations between 1974 and 1977, and are generally much smaller than the "as published" values. Overall, the results for the United Kingdom, Canadian, Australian, Indian and Japanese instruments are very satisfactory. It is worth noting that the No. 83 reference instrument, and the No. 77 Canadian instrument were calibrated independently of each other at Mauna Loa Observatory in 1980, and that the resulting calibrations were found to be "virtually identical" (W.D. Komhyr, personal communication).
TABLE 7.4 Boulder 1977 intercomparison: Approximate average ozone corrections in percent, calculated (a) for the results as published, and (b) with inadvertently imposed reference error component removed from those instruments (*) present at the Belsk 1974 intercomparison (see text for further explanation). Approximate mean ozone correction, % (a) As published (b) Component removed Instrument Country A C AD CD A C AD CD 41* U.K. 4.4 8.5 1.0 1.2 1.1 2.6 1.4 -2.5 71 G.D.R. -5.7 -18.4 6.3 9.2 - - - - 77* Canada 4.3 7.0 -1.1 -8.1 1.0 1.1 -0.7 -11.8 83 U.S.A. ----- Reference -------- ------ Reference ------- 96* Egypt -10.1 -13.6 -2.6 7.2 -13.4 -19.5 -2.2 3.5 105 Australia 0.8 -3.3 2.8 -0.5 - - - - 108* U.S.S.R. 7.7 10.8 -0.7 -13.4 4.4 4.9 -0.3 -17.1 112* India 3.3 8.0 0.2 2.1 0.0 2.1 0.6 -1.6 116 Japan 0.3 0.7 -0.6 -1.9 - - - - Table 7.4 also illustrates the variation in calibration quality that exists among the individual bandpairs and their combinations. The CD band combination has large corrections owing principally to the small CD ozone absorption coefficient, which confers upon this combination an inherently lower accuracy. In practice, the CD combination is usually only used at higher average airmasses and this results in a reduction of the effective mean corrections, perhaps to about half of those shown in Table 7.4. It should be borne in mind that the mean AD corrections tell only part of the story and do not necessarily represent the generally larger corrections for other band combinations or the relative overall quality of the instruments' calibrations. Also, they give no direct indication of the airmass dependence of the corrections or the intrinsic air mass dependences of the instruments due to such things as stray light. It is reported in Peterson (1978) that light scattering effects within instruments 71 and 96 are most pronounced. Possibly this is the main source for the relatively large corrections shown for these instruments in Table 7.4.
The AD results for a large number of other intercomparisons, mainly against instrument No. 83, at Boulder, are reported in WMO (1982), and show generally similar characteristics to those of Table 7.2. Of particular interest are the calibration histories of instruments which have been calibrated more than once, over periods which in some cases extend to twenty years. If the calibration stability of these instruments is summarised by the magnitude of the maximum calibration correction following the initial calibration, then of the nine instruments intercompared more than once with instrument No. 83, two instruments had maxima of less than 1%, a further three had maxima of less than 2%, a further two had maxima of about 3%, and the remaining two had maxima of about 4%. There appears to be no systematic reduction in the magnitudes with time, which suggests that the quality of these particular calibrations has remained stable over the period. The calibration histories each tend to show corrections of the same sign, indicating systematic drifts in instrument calibration with time. Seven of the nine instruments show predominantly negative corrections. This is significant in that the systematic calibration drifts, and the resulting fictitious trends in uncorrected ozone data, will affect large regions, and hence may give the appearance of real ozone trends.
Some difficult questions arise with the retrospective application of calibration corrections. Should the mean corrections be applied as constants to all data, or should the airmass dependences of the corrections be accounted for? Should an attempt be made to separate the correction component due to extraterrestrial-constant error from other components with different functional dependences, such as that due to wedge calibration error? Should the corrections be considered as fixed in time, or be weighted in some retrospectively declining fashion? Should any attempt be made to correct the archives of the World Ozone Data Center? Should the responsibility for making corrections lie with the Center or with the national or station authorities? Some general rules may be formulated in due course, but ultimately the answers to the questions will depend on the characteristics of the particular instrument being considered. Further discussions of these problems are to be found in WMO (1982).
An important issue is the weakness in the current practice of relying on the one instrument (No. 83) as the measurement standard. This is underlined by the presence of the sign error in the calibration of the reference instrument No. 83 between the Belsk and Boulder intercomparisons. Where measurement standards are embodied in reference instruments, it is common, and very desirable, to maintain a group of three or more reference instruments and to represent the measurement standard by the mean result of the largest subset of the group which is operating satisfactorily. This approach allows the prompt detection of any error drift or malfunction in any of the reference instruments and it preserves the measurement standard in the event of any accidental loss of or damage to a reference instrument.
On these grounds it is argued that two additional reference Dobson instruments should be established, by upgrading and independently calibrating two existing instruments to the same high quality as No. 83. Some national reference instruments may already be very close to the quality required. Regular intercomparisons between the three designated reference instruments would be required, and at least two of them would be needed to represent the standard at each international intercomparison of working instruments.
The determination of extraterrestrial constants forms only part of the on-site calibration work. In addition, there are a number of more routine tests which need to be carried out on a regular basis. These comprise:
- the mercury lamp test, which checks the instrument's wavelength calibration and its temperature dependence;- the standard lamp test, which checks the stability of the instrument components of the extraterrestrial constants;
- the wedge calibration test, which checks the density gradient of the optical wedge;
- and the sensitivity test, which checks the overall sensitivity of the instrument.
Descriptions of the tests, with full instructions, are given in the WMO Operations Handbook (Komhyr, 1980b).
Errors can arise if there are instrument drifts and the tests are neglected, or if the tests are improperly done, improperly interpreted, or improperly documented. As was noted earlier, it is not clear to what extent the detailed instructions in the WMO Operations Handbook are rigorously followed throughout the network. The comparisons given in Section 5.4 between calculated and measured standard errors in standard lamp readings indicate that there is still room for improvement in the use of the standard lamp test. Standard lamps can also age with use and a set of lamps must be maintained to guard against this. The errors due to the set of tests and their execution are difficult to estimate, but might amount to a few percent, if gross error is excluded. Olafson and Asbridge (1981b) report the curious discovery that some of their mercury lamps appear to give inaccurate wavelength calibrations. This certainly needs further study.
Recently, the Central Dobson Spectrophotometer laboratory has begun circulating reference standard lamps around the network. The application of a similar lamp, UQ1, to 15 instruments which had just been calibrated by direct intercomparison yielded a standard deviation of 0.006 in NAD (Komhyr et al., 1981a). This suggests that NAD errors of greater than, say 0.015, equivalent to XAD mean errors of about 2%, could be corrected on the basis of these lamp measurement
The WMO Operations Manual also gives instructions for a variety of electronics and optics maintenance tasks which may be carried out on site, for such things as cleaning the commutator and replacing the photomultiplier. The instructions and advice reflect a considerable accumulation of practical experience and should be followed carefully. Generally, any effects of poor maintenance will show up in the routine tests. The design of the instrument does not call for any precise electrical calibrations.
7.4 Routine operation
Errors in day to day operation can be attributed to either deficiency of judgement or error of execution. In the first category come questions such as: deciding what measurement type to use, e.g., direct sun AD versus zenith sky AD, choosing the null balance when the light is weak and the ammeter is consequently unsteady, deciding when to take a measurement, judging when a measurement is affected by extraneous factors and needs repeating, and deciding what notes, if any, should be recorded. The quality of the judgements affects not only the current day's ozone estimate, but also any subsequent use of the data in constructing empirical charts for zenith sky clear and cloudy estimations. Generally, the quality of judgements are most crucial when the observing conditions are difficult, such as when the solar elevation is low or is changing rapidly, or the weather is poor. Local practises in making these judgements are likely to be varied.
Errors of execution include: errors in aligning the instrument and the sun director; errors in the choice of appropriate accessories; errors in setting the Q plate levers; failure to properly adjust for changing instrument temperature; error in reading the wedge dial; and error in recording the readings. Gross errors, such as confusing the A and C bandpairs, are readily detectable by elementary quality control measures, and in any case are likely to be rare and to have little effect on climatological mean data. Small systematic errors may be a problem though. Particular systematic error sources are:
- the failure to keep the ground quartz plate diffuser clean and free of UV absorbing oils;- the failure to minimise, and frequently adjust for, the temperature changes of the instrument (see Section 2);
- and the failure to refresh the silica gel dessicant sufficiently often.
Individual operators may exhibit systematic biases in setting the Q levers and in reading the wedge dial, but the resulting errors should be fairly small, i.e. less than 0.5%.
The precision of the Dobson instrument under normal circumstances is very good, with standard errors being small compared to potential systematic errors. The standard error in single routine observations is of the order of 0.5% in the mean AD ozone estimate, according to the work of Dobson and Normand (1962) and to the analysis of wavelength band uncertainty in Section 5. Occasionally it may rise to 1%. Dziewulska-Losiowa and Walshaw (1975) found a standard error of about 0.6% from a study of the Belsk intercomparison data.
The AD ozone measurement random standard errors given in NASA (1979, Table 6.2), and thence copied to WMO (1981, Table 2.1) and WMO (1982, Table 3.1), are estimates which include the effects of such things as: short term calibration fluctuations; observer error; varying haze conditions; and the temperature dependence of ozone absorption, and are therefore not the same as the precisions discussed above. The NASA (1979) standard error estimates were 1.5% for optimally maintained and operated instruments and 3% for an average network instrument for direct sun AD measurements, and similarly, 2.5% and 5% for zenith sky observations. These estimates are approximate of course, and there are obvious questions as to how well such "random" standard errors actually represent the variable systematic errors. Overall summaries of error for the present study are given in Section 14.
Special mention must be made of high latitude stations, for most operational difficulties are exacerbated by increasing latitude. The greater airmasses and greater ozone amounts result in low light levels and high differential absorption, both of which tax the capabilities of the instrument and necessitate the use of the less accurate C and CD longer wavelength hand combinations and possibly the focussed image method. During winter, focussed image moon measurements may be the only possibility. The greater variation in the ozone layer's mean temperature (see Section 9) and mean height (see Section 12) at higher latitudes introduces greater uncertainty. Meteorological conditions may bring extensive cloudiness which requires the use of the less accurate cloudy sky estimation methods (whose empirical charts may be difficult to construct at such latitudes). The instrument will be subjected to the stresses of very low temperatures and of rapidly varying temperatures. The stations can be very isolated and the observing work can be very difficult. To some extent, however, these intrinsic difficulties will be offset by the higher standards of operation expected of the well motivated, well trained staff who usually man the high latitude stations.
Sampling errors may be of importance to statistical averages. Day to day variation in column ozone amounts can be large, of up to about 5% in the tropics and about 30% at high latitudes, and so losses of data due to weather or operational conditions may cause biases of several percent in some monthly means, especially since ozone amounts are strongly correlated with synoptic weather conditions (WMO, 1982). Mean diurnal variations are thought to be negligible. Spatial sampling errors also exist, of course, owing to the very uneven spatial distribution of the network. It would seem that further quantitative work on both temporal and spatial sampling errors is needed. Satellite data will be very useful in this respect.
A certain amount can be done to improve the instrument's ease of operation, especially in the more severe climates. For example, in the United States network, the instruments are operated from an out-building which is equipped with a heater, an air conditioner, a rotating instrument table, and a motor driven astronomical dome roof. In Toronto, the instrument is mounted indoors with a periscope system projecting through the ceiling. Westbury, Thomas and Simmons (1981) report an ergonomic design for a null balance display consisting of a ring of sixteen light-emitting diodes. The Q levers can be equipped with permanent, though adjustable, stops at the A and D positions.
The degree of possible automation of the instrument is restricted by the need to make accessory changes to suit the airmass and sky conditions. Raeber (1973) described the partial automation of instrument No. 51, at Arosa, in which the instrument's temperature, Q lever settings, photomultiplier voltage, null balancing and data acquisition were controlled. Komhyr (1982, personal communication) reports a similar development using readily available electronic sub-assemblies which are programmable and can process data, and which can be added to existing instruments with the minimum of interference. Such semi-automated systems relieve the operator of much of the tedium of the operation of the instrument, and they ensure a more efficient and accurate data collection. They are especially valuable for the collection of Umkehr data. It is to be hoped that their use becomes more widespread.
7.5 Summary
(i) The success of the Dobson network is dependent on the continued goodwill and cooperation of the many independent institutions throughout the world who own and operate the instruments.(ii) The instrument is manually operated. Careful attention is needed to the human aspects of its operation. Instructional material, and therefore actual practice, has varied in the past, but the new WMO Operations Manual (Komhyr, 1980b) should gradually unify practices throughout the network. It would be desirable to institute a comprehensive quality assurance programme.
(iii) The accuracy of extraterrestrial constants independently determined under particularly clear and stable atmospheric conditions can be equivalent to an accuracy of better than 1% in the mean AD ozone estimates. However, under more usual conditions, a figure of 5% or more is common. The resulting ozone errors vary inversely with airmass.
(iv) Extraterrestrial constants for network instruments are best obtained from direct intercomparisons against well-calibrated reference instruments. Successive intercomparisons over many years have shown that AD calibrations can be maintained to within 3% (in mean ozone) for most instruments, and to within 1% for perhaps a third of the instruments.
(v) Errors were present in the calibration of the World Reference Dobson Spectrophotometer No. 83 between 1972 and 1976, but the effect on the standard AD ozone estimation was only about 0.4%. The errors for other band combinations amount to several percent however. The errors were transferred to the instruments calibrated against No. 83 at the 1974 Belsk intercomparison. The 1977 Boulder intercomparison results implicitly include corrections for these errors in the Belsk-calibrated instruments present.
(vi) The present reliance on only one primary reference instrument is unsatisfactory. A further two instruments should be upgraded to the primary reference quality and the mean result of the three instruments taken to represent the World Standard.
(vii) There are a number of important problems yet to be resolved in how to best apply to archived data the corrections deduced from intercomparisons.
(viii) Routine (usually monthly) tests allow the detection of calibration changes equivalent to errors of a few percent in mean AD ozone estimations. Travelling standard lamps can ensure a network consistency in extraterrestrial constants equivalent to 2% in mean AD ozone.
(ix) The precision or standard error, of individual AD ozone measurements is in the 0.5% to 1% range for normal circumstances. Systematic errors associated with routine operation, such as due to instrument temperature drifts or poor choice of measurement type, might add a further 1% to 3% error, depending on the circumstances.
(x) Most systematic error sources become much greater at higher latitude sites, owing principally to the lower light levels, higher differential absorption, and more difficult observing conditions encountered there. Less accurate estimation methods, e.g. CD direct sun, often must be used.
(xi) Uneven temporal and spatial sampling may result in errors in some average data statistics of up to a few percent on occasion. More study of this problem is needed.
(xii) The development and use of systems which semi-automate the Dobson instrument should be encouraged. Such systems improve the quality and quantity of data collected, and are especially valuable for Umkehr measurements.
Return to Table of Contents
Forward to Electronics