SPARC logo (45 Ko)
S P A R C

Stratospheric Processes And their Role in Climate
A project of the World Climate Research Programme

Home Initiatives Organisation Publications Meetings Acronyms and Abbreviations Useful Links

 

SPARC/IOC/GAW

Assessment of Trends in the Vertical Distribution of Ozone

 

2.5. Ozonesonde Analyses

2.5.1. Introduction

In the previous sections 2.3.3.2 and 2.4.1, time series of ozonesonde measurements obtained by various sounding stations were compared with SAGE I and SAGE II respectively to investigate differences or drifts between the different long-term time series of ozone measurements in the lower/middle stratosphere. The emphasis of this section is on comparison of ozonesondes with other non-satellite ozone measuring systems to quantify the precision and accuracy of the three different types of ozone sondes, i.e. Brewer-Mast (BM), Electrochemical Concentration Cell (ECC), and the Carbon Iodine (KC79), that are in routine operation at the various sounding stations (see chapter  1).

Since 1970 dedicated short term campaigns to intercompare the different types of ozone sondes used in operational networks have been carried out at different times during the course of the ozone sonde record. These intercomparisons are intended to assess the performance of and to quantify any systematic differences between the various types of sondes. Most of the campaigns have been conducted in the field where at least two ozone sondes were carried to altitudes of about 30 to 35 km by the same balloon. The WMO Jülich Ozone Sonde Intercomparison Experiment (JOSIE) held in 1996 is the first intercomparison campaign that compared all operational types of ozonesondes in a controlled environmental chamber capable of simulating real flight conditions (Smit et al., 1997). The controlled environment plus the fact that the ozonesonde measurements could be compared to an accurate UV-Photometer as a reference allow the opportunity to carry out experiments to address questions arising from the previous field comparisons. However, short term intercomparisons are more or less "snap shots" and may not necessarily reflect the performance of ozonesondes under operational field conditions. Comparison studies of time series of ozonesonde data with other simultaneously operating ozone monitoring devices like lidar or microwave are more valuable to assess the data quality of the ozonesonde measurements in regular operation.

In the free troposphere ozone soundings provide the only time series of measurements to derive long term ozone trends. Due to the much lower concentrations of ozone in the troposphere compared to the stratosphere the performance of the sondes and their typical instrumental/operational factors determining precision and accuracy are different in the two regions of the atmosphere. In this section special attention will be paid to how much confidence exists in the data quality of the different ozonesonde records available for ozone trend assessment in the troposphere.

 

2.5.2. Laboratory Studies: Jülich Ozone Sonde Intercomparison Experiment (JOSIE)

In February/March 1996 JOSIE was conducted in the environmental simulation chamber at the Research Centre Jülich (Germany) to assess the performance of the major types of ozonesondes (ECC , Brewer-Mast , KC79) used within the Global Atmosphere Watch (GAW) and Global Ozone Network (GLONET). Eight ozone sounding laboratories participated in the intercomparison effort. Four ozonesondes were "flown" simultaneously and compared to a UV-photometer as a reference, while pressure, temperature and ozone concentrations were regulated to simulate different types of vertical soundings with an ascent velocity of about 5 m/s up to 35 km altitude (Smit et al., 1994).

Four different types of ozonesondes were "flown" simultaneously covering six simulated ascents (four mid-latitude and two tropical profile types). Quantitative results of the sonde comparisons with the UV-photometer are shown for each participating sounding laboratory in Figures 2.42a to 2.42h. The comparison is presented as relative deviation of the sonde readings from the UV-photometer (typically within ±10-20%). However, some systematic deviations were observed with increasing magnitude in the middle stratosphere for some participants (CMDL [2], JMA [8]). Apart from systematic deviations, there are precision errors of varying size for the different participating laboratories. The best precision (±5%) was achieved by CMDL [2], KFA [3], AES [6] and CNRS [8] (excepted one outlier profile) which were operating the ECC-type ozonesondes. Conversely, the non-ECC types of ozonesondes, operated by MOHP [1], IMD [4], ASP [5] and JMA [8], exhibit a somewhat lower precision of about ±10-15%. The relative precision is best for all sondes in the middle stratosphere where ozone is at a maximum. The non-ECC type sondes show a considerably lower precision in the troposphere and stratosphere compared to the ECC sondes.

 

Figure 2.42 JOSIE : Relative deviations of the individual sonde readings from the UV-photometer for each participating laboratory obtained from the six simulation runs of the mid latitudinal (solid lines) and the tropical profiles (dotted). Tropospheric part of tropical profiles below 20 km are excluded. ECC-SPC5A and ECC-SPC6 are manufactured by Science Pump Corporation and ECC-ENSCI by ENSCI-Corporation. BM-orig. is the original model of the Mast-Company, BM-hybrid is a hybrid of an ozone sensor from Mast-Company coupled with a teflon pump from ENSCI-Corporation. The Indian sonde is made from a Brewer-Mast like ozone sensor and a teflon pump manufactured by the Indian Meteorological Department. Jap.KC79 is the Japanese KC79 sonde.

 

Table 2.8 shows the total ozone column normalisation factor, which is the ratio of the integrated ozone profile measured by the UV-photometer and each sonde, for each sonde "flown" plus the average factor and standard deviations for each participating laboratory. Although all individual normalisation factors range between 0.9 and 1.1, the variability (standard deviation) of the ECC-sonde types with values of ±0.02 for three of the four ECC-stations are significantly smaller than the variability of the other types of sondes with values of ±(0.07-0.10). In the case of the CNRS [7], the standard deviation of 0.07 is mainly caused by the relative low normalisation factor obtained from the first simulation run in campaign II (see Table 2.8 and Figure 2.42G).

 

 

Sonde Type / Participant [No]
Total Ozone Normalisation Factor individual simulation runs
Average
Standard Deviation
BM (original) / MOHP [1]
0.95
1.03
1.01
1.13
0.94
0.94
1.00
0.07
ECC (ENSCI) / CMDL [2]
0.93
0.92
0.90
0.88
0.93
0.92
0.91
0.02
ECC (SPC6A) / KFA [3]
1.01
0.98
0.97
0.98
1.00
1.01
1.00
0.02
Indian / IMD [4]
1.06
(*)
0.97
1.05
0.90
1.06
1.01
0.08
BM-Hybrid / ASP [5]
1.03
1.01
0.91
1.06
1.22
1.05
1.05
0.10
ECC (SPC5A/ENSCI) / AES [6]
0.96
0.97
0.94
(*)
0.97
0.94
0.96
0.02
ECC (SPC5A) / CNRS [7]
0.90
1.09
1.03
1.09
1.07
1.07
1.04
0.07
KC79 / JMA [8]
0.86
0.96
0.84
0.89
1.09
0.85
0.91
0.10

Table 2.8 : JOSIE: Summary of total ozone column normalisation factors. Factors marked by (*) are non valid sonde data. Total ozone normalisation were only applied by the following stations: MOHP [1] , IMD [4] , ASP [5] , AES [6] and JMA [8].

 

Even if the total ozone normalisation (correction) factor is not used to correct the sonde profile it provides an excellent screening test for suspect soundings. However, the normalisation factor is not a guarantee that the profile is correct. Trends or discontinuities in normalisation factor time series can also be a cause for concern (see Chapter 3).

The results of JOSIE presented here, show that in general the sondes agree well with the ozone reference. Most sondes track the simulated ozone profile quite well, even under extreme low ozone concentrations in the tropical troposphere. The JOSIE results show that the ECC-type sondes perform better than other sonde types with regard to precision and accuracy. The observed differences are mostly due to differences in the preparation and correction procedures applied by the different laboratories. The performance of the non-ECC type sondes is quite different as are the observed deviations from the UV-photometer measurements. The results of JOSIE also demonstrate that there is an urgent need for the homogenisation of operating procedures. Further, the larger variability among ozonesonde measurements in the middle stratosphere is due mainly to greater uncertainties in the pump efficiency and the amount of sensing solution that has evaporated.

The increasing variability above 25 km altitude, particularly observed in the case of the non-ECC sonde types, is mostly caused by the decaying pump efficiency and its increasing uncertainty at lower pressures. At the present the pump flow rate is typically corrected at lower pressures by using an average pump efficiency measurement curve (see chapter  1). More recent measurements indicate that the pump efficiency is smaller than the average by about 10-12% at 5 hPa for the ECC-Sonde (Johnson et al., 1997) and by even larger amounts for the Brewer Mast sonde (Steinbrecht et al., 1997; De Backer et al., 1997). It should be noted that significant differences (more than 5-10% at 5 hPa) are observed between the various experimental methods used to determine the pumpflow efficiency at lower pressures.

An additional source of uncertainty is that by the time an ozonesonde reaches the middle stratosphere it has been operating for nearly 90 minutes and at this stage of the flight the uncertainties in the sensor cell characteristics are greater. First, a certain percentage of the sensing solution has evaporated at a rate dependent on the temperature of the cell and ambient pressure encountered during flight. For the ECC-sondes this means that due to evaporation, the concentration of the sensing solution increases, which can have an enhancing effect on the sensitivity of the ECC-sensor and thus on the measured ozone (Barnes et al., 1985; Komhyr, 1969).

There are some experimental indications for ECC-sondes, that if the preparation and correction procedures prescribed by Komhyr (1986) are used, this sensitivity enhancement effect can compensate for the too low conventional pumpflow correction in the middle stratosphere. During JOSIE, this may be the case for the ECC-sondes operated by KFA [3], AES [6] and CNRS [7] which were using the same initial concentrations of sensing solutions and the same pumpflow corrections (Komhyr, 1986) and did not show any systematic deviations from the UV-photometer in the middle stratosphere. This in contrast to the overestimation above 25 km altitude by the ECC-sondes operated by CMDL [2] which is probably caused by the larger pumpflow correction (based on individual pumpflow efficiency measurements in the laboratory) applied in combination with the sensitivity enhancement of the ECC-sensor due to evaporation. However, the process is not understood in detail up to now and more investigations are necessary to study this particularly in relation to the initial KI-concentration, the actual temperature of the sensing solution and the pumpflow correction efficiency which also has an important influence on the sonde performance.

The uncertainty in sonde measurements of tropospheric ozone is relatively high since the signal-to-noise is low. The impact of instrumental errors is larger when measuring the much lower values of tropospheric ozone. One instrumental error is uncertainty in the sensor background current which varies in magnitude from one sonde to another as well as from one sonde type to another. Correction for the background current can have a significant impact on the measured tropospheric values in regions where the ozone concentration is low, i.e. near the tropopause. For ECC-sondes the conventional method of correction prescribed by Komhyr (1986) assumes the background current is dependent on the oxygen partial pressure and decreases with altitude. Several laboratory studies (Thornton et al., 1982; Smit et al., 1994) do not show any oxygen dependence on the background current and the accuracy of the ECC-sonde is significantly improved by using a constant background current correction throughout the entire vertical profile.

Recent laboratory studies of ECC-sondes have also shown that the background signal can be correlated to past exposure to ozone. As a consequence, the background current should be measured before exposure to ozone in the preparation procedure as was shown in a field study by Reid et al. (1996). No background correction is made to the BM-sonde records, but prior to flight the BM-sonde readings are electronically compensated for the background current. Nothing is known about the fate of the background of the BM-sonde during flight. Any changes in the magnitude of the background current over the sonde record will directly affect the trends derived for the free troposphere. It is important that more research be dedicated to the size and impact of the changes in the background current of different sonde types.

Another source of uncertainty is the influence of local air pollution which can have detrimental effect on the sonde performance. Most known is the negative interfering effect by SO2, which lowers the ozone readings and can have a memory effect if excess of SO2 is accumulated in the sensing solution of the sensor (Schenkel et al., 1982). At Uccle the SO2 contamination was recognised as a serious problem in the lower tropospheric part of the sounding profile and a procedure to correct for the SO2 interference has been developed (DeMuer et al., 1993). It may be possible that in the early 1970’s when the SO2 concentrations were much higher than in the 1980’s and 1990’s, the records of other stations like Hohenpeissenberg, Tateno and Sapporo are also affected by SO2 interferences (Logan, 1994).

 

2.5.3. Dedicated Short Term Ozonesonde Intercomparison Campaigns

During the course of developing the ozonesonde record, several dedicated, short term intercomparison campaigns in the field were conducted. A summary of the campaigns, which started in 1970, is given in Table 2.9.

 

Table 2.9 : Summary of ozone sonde intercomparisons between 1970 and 1996. Dates and locations of the intercomparisons are shown and the types of profiling instruments with a reference are indicated.

 

The earlier campaigns included only ozone sondes so comparison of sonde results to a reference profile measured by a separate "standard" technique is not possible. This type of comparison allows the determination of relative errors between the different sonde types as a function of altitude but not absolute systematic errors. However, all comparisons normalised the profiles to a common ground-based total ozone measurement to minimise systematic differences. Later intercomparisons (BOIC, MAP/GLOBUS, STOIC, OHP I & II, MLO3, JOSIE, GAP) used a reference profile measured by other techniques and results of these comparisons yield better estimates of absolute errors for the sonde measurement as a function of altitude. A significant amount of information has been learned from these intercomparisons and a review of results from the campaigns has provided information regarding the confidence with which the sonde data may be used for long term trend analysis. The campaigns offer the opportunity for operators to compare flight preparation procedures and to evaluate the effects of any differences.

Results of the intercomparisons indicate that sonde measurements made in the lower stratosphere (12 to 27 km) are quite reliable for use in long-term trend analyses. The systematic differences between different sonde types have been less than 5% and the random variability from one sonde to another for all sonde types has been less than 5%. This relatively low variability is a result of the normalisation of the profile to ground-based total ozone measurements (see Chapter 1). The normalisation is mainly weighted to the ozone in the lower stratosphere which contains most of the column ozone.

The intercomparison projects have shown that the variability among ozonesondes generally increases again in the middle stratospheric region (27 km to balloon burst altitude). For example, in the BOIC campaign 3 ECC ozonesondes (from different groups) and 1 Brewer-Mast sonde showed differences of about 10% as the balloon approached 30 km altitude. Results of the WMO III Vanscoy intercomparison included two additional types of sondes, the Indian and Japanese KC79 sondes. The variability in comparing all 4 types of sondes was even higher at about 10-15%, reaching 20% at 5 hPa (~35 km). Radiosonde pressure variability adds to the uncertainty in ozone profiles in the middle stratosphere, which was observed in the Vanscoy (WMO III) intercomparison. The magnitude of variations and errors in ozonesonde measurements in the middle stratosphere have been fairly consistent in showing larger variability in this region from past intercomparison projects and individual tests, thus indicating that trend analysis may be less reliable at these altitudes. However, recent results from the JOSIE campaign have shown very good precision and accuracy extending into the middle stratosphere, especially for the ECC type sondes.

The issue of tropospheric ozone measurement accuracy is addressed fully in the next section, but there are a number of results from the intercomparison exercises, which can have a bearing on trend determinations in the troposphere. Results of the intercomparisons show systematic differences between sonde types which are typically 10 to 15% but can be as much as 25% for some campaigns. Most campaigns (e.g. WMO I, WMO II, BOIC, OHP I, OHP II) indicate that the ECC sonde measures about 15% to 30% more than the Brewer-Mast in the troposphere. However, results of WMO III suggest that the Brewer-Mast measures about 15% more than the ECC in the troposphere. This marked change over time , with the BM-sonde lower than the ECC in 1970 and larger in 1991 may indicate that any determination of the long term ozone trend in the troposphere is subject to considerable uncertainty, particularly when different ozonesonde types are been used for different parts of the record.

Like the JOSIE-campaign findings, the field intercomparisons have shown that in the troposphere the sonde measurements are less precise (typically about ± 15%, 2-sigma) than in the stratosphere. In general the Brewer-Mast type sondes are less precise than the ECC sondes. One outcome of the intercomparisons is the discovery that occasionally sondes measure erratically (errors >± 50%) in the troposphere with no apparent explanation. The tropospheric error would not be apparent if these profile measurements were considered on their own.

It should be noted that the short term campaign results may not necessarily reflect the performance of ozone sondes under operational field conditions. Particularly for the non-ECC-type of sondes, like the older established BM-sounding stations (Hohenpeissenberg, Payerne and Uccle), the in-flight performance characteristics appear to be strongly coupled to the operational procedures followed at the different stations, but also to the location of the launch site. In the next sections the performance of the different sonde types is evaluated through comparison of time series with other ozone profiling techniques in regular operation at the sounding station site.

 

2.5.4. Ozonesonde Comparison Studies in the Troposphere

Identification of trends in tropospheric ozone from the ozonesonde record demands that the measurement accuracy of the sondes remains constant over a long time. As already mentioned, from the WMO intercomparison campaigns it appears that large relative changes in accuracy between BM and ECC sondes may have occurred between 1970 and 1991. However, as will be shown in this section, there is also strong evidence that the ECC and BM records (at least at Hohenpeissenberg) are more reliable than results from the WMO intercomparisons would imply.

The ozone profiling capabilities of the ECC-sonde during regular operations were evaluated in a study by Ancellet and Beekmann (1997) through comparison with routine lidar measurements made at the Observatoire de Haute Provence in the 1990-1995 period (Ancellet et al., 1989). The seasonal means of mid tropospheric ozone (4.5-5.5 km altitude) obtained by the sonde and lidar are summarised in Table 2.10. ECC values are not corrected by total ozone normalisation. Both data sets show excellent agreement in the annual mean (54.3 ppb for ECC, 53.8 ppb for Lidar) and in seasonal variations in the mid troposphere. Differences in particular periods are generally in the range of 2 ppb (5%). For 15 simultaneous ECC versus lidar profiles the mean of the differences observed between 4 and 7 km was 2.5± 1.8 ppbv (4± 3%) which is in good agreement with differences between the seasonal means (see Table 2.10). This comparative study indicates that tropospheric ozonesonde profiles measured with ECC-sondes should not be corrected for the total ozone column. Supporting evidence exists from simulation experiments performed at the ozonesonde calibration facility in Jülich (Smit et al., 1994). Although there is a dearth of comparative studies to evaluate the performance of ozonesondes in the troposphere, this present study under regular field operations together with the studies under controlled laboratory conditions show that there is sufficient confidence in the performance of the ECC-sondes to use them for tropospheric trend assessment studies.

 

Instrument
ECC / OHP
Lidar / OHP
Period
1991-95
1990-95
Time
mean
n
mean
n

Nov.-Feb.

49.0 ± 0.8
66
47.2 ±1.2
96

March

53.5 ± 2.7
18
53.8 ± 2.7
23

Apr.-Jun.

65.2 ±1.5
52
63.2 ±2.3
56

Jul.-Sep.

63.0 ± 1.5
54
60.7 ± 1.8
74

October

53.1 ± 2.1
52
55.2 ±3.0
14

all

54.3 ± 0.6
207
53.8 ± 1.2
263

Table 2.10 : Seasonal mean ozone values and the standard deviation (in ppbv) of the mean in the mid-troposphere (4.5 -5.5 km height) at the Observatoire de Haute Provence; n denotes the number of profiles.

 

As mentioned before, one major uncertainty in BM profiles is the pump flow correction at lower pressures (<100 hPa). Individual BM pumpflow calibrations in the laboratory [De Backer et al., 1997, Steinbrecht et al., 1997] have shown that at lower pressures the pumpflow correction should be significantly larger than the current correction recommended by the WMO [Claude et al., 1986]. Although the correction itself is only applied to the upper part of the profile, the additional act of normalising to the total column measurement means that erratic pump corrections will also show up in the tropospheric part of the profile.

The effect of the pump correction factor on the tropospheric performance of BM-sondes in combination with the use of the total ozone normalisation was investigated at Uccle using both ECC and BM sondes on the same balloon. For each BM-sonde flown, the pump efficiency was individually measured in the laboratory before launch. The results from 20 dual sounding are shown in Figure 2.43. When the BM sondes are corrected with the standard WMO pump correction, the BM-sondes measure about 10% more ozone in the troposphere than the ECC-sondes and 3-5% less ozone above 20 km altitude. However, when using the individual measured pump efficiency corrections, the differences between BM and ECC soundings are within ± 2% and show no significant altitude dependence. It should be noted that the rms difference of 8-10% between the ECC and BM measurements in the troposphere is consistent with the results obtained during JOSIE (see section 2.5.2). This study shows that the standard pump correction recommended is inadequate for the BM-sonde and because of the total ozone normalisation, it can have a significant effect on tropospheric ozone values.

 

Figure 2.43 Mean percentage differences between ozone profiles obtained from measurements with BM and ECC sondes with the standard (grey) and the new pump (black) corrections in Uccle. The domains where the differences are statistically significant at the 95% confidence limit, according to a 2-tailed student-t test are indicated by thick lines. The dashed lines give the 1 sigma levels of the differences.

 

This issue can be relevant for trends if the recently measured pumpflow corrections, which are larger than the standard correction recommended by WMO, has been caused by a decrease in the BM-pump efficiency (at lower pressures) during the sonde record. However, at present this is not clear and the observed differences may be caused by differences between the different experimental methods used to measure the pump efficiencies at low pressures. Unfortunately, the standard pump flow corrections recommended for both the BM-and ECC-sondes are based on rather old pump flow efficiency measurements which were performed more than 25 years ago and they never have been officially re-evaluated in the literature. Therefore, it is rather difficult to trace any long term changes in the pumpflow efficiencies at lower pressures during the sonde record.

The previous study at Uccle also suggests the need to reconsider the concept of applying the total ozone normalisation factor to the tropospheric part of the ozone profile measured by the BM-sonde. Results from comparison studies (Thouret et al., 1997; and Jeker, private communication) of BM-soundings at Hohenpeissenberg and Payerne with ozone measurements made from aboard in-service aircraft suggest that it is indeed more proper not to use the total ozone normalisation factor in the tropospheric part of the sounding. Similar findings were obtained from the BM/ECC-comparison flights during SONDEX 96 (Schmidlin, private communication). However, this issue is still in an explorative stage and more research is required.

To evaluate the long term performance of the BM ozone soundings at Hohenpeissenberg the sonde record obtained at 700-800 hPa was compared with night-time measurements made at the Zugspitze mountain station at 2970 m altitude (700 hPa). Night-time measurements at this site are generally indicative of free tropospheric conditions and they have been made with a UV-photometer or a chemiluminescent instrument since 1978. The results of the comparison is shown in Figure 2.44. Although there is an offset between the measurements (800 hPa data from the sondes agrees better with the Zugspitze data than the 700 hPa measurements), there is no major anomaly in the long term variations measured by the sonde and at the Zugspitze. At both stations the strongest increase in ozone mixing ratio was not observed until the beginning of the eighties. Since then no significant trend has been observed. These observations strongly suggest that the trend in tropospheric ozone derived from the Hohenpeissenberg data is reliable to about 0.3% per year and it is obviously not affected by uncertainties induced via the pump flow correction in combination with total ozone normalisation. However, the offset between Hohenpeissenberg 700 hPa and Zugspitze data (comparable altitude) might be due to regional variations in the vertical ozone distribution, but it could also be caused by the total ozone normalisation leading to an 8-10% shift in the tropospheric data or from other non identified offsets.

A similar comparative study has been done for the 700 hPa level of the Payerne BM-ozone soundings which were compared with the data sets of 2 high alpine surface monitoring stations (Jungfraujoch and Zugspitze) to identify the anomalies observed between 1990 and 1993 (see Figure 2.45). Since 1992, very good agreement between the annual ozone concentrations at Zugspitze and Jungfraujoch has been found, and the deviations relative to the sounding data remained approximately constant with time before 1990 and after 1993. The small temporal drift in the difference "sounding-Zugspitze" has been further reduced by applying the results of the pre-flight laboratory calibration of the ozone sondes (see chapter  1). However, it should be noted that the 1982/1983-period the Payerne sonde record show inconsistencies in the troposphere which cannot be explained and are probably not of atmospheric origin (Staehelin, private communication). Therefore, the ozone record over this period has to be taken with caution to derive tropospheric ozone trends.

 

Figure 2.44 Ozone trend component based on monthly means of Hohenpeissenberg Brewer Mast data at 700 and 800 hPa (approx. 3000 and 2000m a.s.l.) and continuous measurements at the Zugspitze (2960 m a.s.l.) since 1978 [courtesy of K.E. Scheel].

 

Figure 2.45 Time series as 12-months moving averages of the monthly means of ozone [ppb] of the Payerne sounding (700 hPa, ~ 3000 m asl) and of the surface monitoring stations of Jungfraujoch and Zugspitze between 1984 and 1996. [Zugspitze-data by courtesy of Dr H. E. Scheel, Jungfraujoch-data by EMPA, Switzerland].

 

2.5.5. Ozonesonde Comparison Studies in the Stratosphere

The results of the short term intercomparisons have in general shown that the sondes are giving consistent results in the lower/middle stratosphere between 12 and 27 km provided the profiles are normalised to a ground based total ozone measurement. The different sonde types have consistently agreed to within 5% for this height range. The results are not as conclusive for altitudes above about 27 km. In this section the conclusions drawn from the "snap shot" intercomparisons are evaluated through comparison of individual sonde records with an independent ozone profiling technique done under regular operation at the sounding site.

At a few sounding stations over the last 5-10 years, other vertical ozone profiling techniques like lidar and/or microwave instruments, have been used to monitor the vertical ozone distribution. These data have been used in this assessment to validate the different types of sondes in regular operation. The sounding sites, sonde type, "reference" instrument and period of comparison are listed in Table 2.11. All installed instruments are dedicated to stratospheric profiling as part of the Network for the Detection of Stratospheric Change (NDSC). The lidar systems listed have been also used in this assessment to validate SAGE II data (section 2.4.2), and the microwave radiometer has been used in a similar way at Lauder (section 2.4.8).

 

Comparison Specifications
[(Sonde-Ref)/Ref] x 100%
Sounding Site
Sonde Type
Comparison Instrument
Observation Period
Number Pairs
15-20 km
20-25 km
25-30 km
30-33 km
Lauder ECC
Lidar
11/94-12/96
» 80
-2± 2
2± 2
4± 2
3± 2
Lauder ECC
Microwave
11/92-12/95
» 130
 
-2± 2
-1± 2
-1± 3
Haute Provence ECC
Lidar
11/91-12/96
» 55
2± 4
3± 2
2± 3
-3± 3
GSFC Mobile Station ECC
Lidar
10/88-08/95
» 90
6± 5
4± 2
3± 1
1± 2
Hohenpeissenberg BM
Lidar
11/87-11/95
(*)
4± 3
3± 2
-3± 3
-10± 5
Payerne BM
Microwave
11/94-09/96
» 300
-3± 3
-2± 2
-5± 5
 
Haute Provence BM
Lidar
07/86-12/90
5
 
5± 5
-2± 8
 
Tsukuba KC79
Lidar
11/88-05/91
» 45
 
-2± 4
3± 3
6± 4

Table 2.11: Summary of ozonesonde comparisons with other ozone profiling techniques in the stratosphere. Pairs marked with asterisks (*) are based on differences between monthly means of sonde and lidar data

 

Time series of the differences between the sonde record and the co-existing "reference" instrument were calculated for each site at altitude levels between 15 and 35 km with an altitude step of about 2.5 km. The time coincidence criterion between sondes and microwave measurements was ± 24 hours. Differences between the monthly means of the sonde and the lidar were obtained only for the BM-Sonde/Lidar comparison at Hohenpeissenberg. The vertical profiles of the mean relative differences (± 2 standard error) between the different sondes compared with the co-existing "reference" are shown in Figure 2.46 for ECC sonde comparisons and in Figure 2.47 for the non-ECC sonde comparisons. In addition, the major results of the comparisons are summarised in Table 2.11.

The ECC sonde (Figure 2.46) shows agreement with lidar or microwave data to better than ± 2.3% for individual comparisons throughout the lower/middle stratosphere up to the balloon burst altitude at 33-35 km. Averaged over the three comparisons with lidar the differences are ~2± 3% at 15-20 km, 2± 2% at 20-25 km, 2± 2% at 25-30 km, and -1± 3% at 30-33 km. It is obvious that the ECC-sonde tends to slightly overestimate the ozone compared to the lidar, even in the altitude region above 25 km where the conventional pump correction is too low. This is probably caused by the compensating effect of the ECC-sonde sensitivity increase due to an enhancement of the concentration of the sensing solution by evaporation during the course of the sounding (see section 2.5.2). This effect even slightly dominates up to an altitude of 30 km. The small negative differences obtained from the comparison with the microwave at Lauder can be also caused by the fact that the microwave instrument slightly overestimated ozone during the observational period (see microwave SAGE II comparison in section 2.4.8).

 

Figure 2.46 Average differences between ECC sondes and lidar or microwave radiometer as a function of altitude at a ± 24 hours coincidence criterion. The error bars are ± two sigma standard error.

 

The long term comparison between BM sonde and lidar data at Hohenpeissenberg is shown in Figure 2.47 (upper left panel). Between 15 and 28 km altitude the sonde readings are about 3± 3% larger than the lidar readings. Above 28 km altitude the BM-sonde readings are significantly decreasing with altitude compared to the lidar (-6% at 35 km). This is due to the deteriorating pump efficiency of the sondes, which is not properly corrected by the current WMO procedure [Steinbrecht et al., 1997]. However, at present there is no evidence that this problem can cause any trend effects in the sonde record at Hohenpeissenberg.

BM sondes at Payerne (see Figure 2.47, upper right panel) underestimate ozone in the lower stratosphere by about -3± 3% compared to microwave measurements. Above 28 km the sonde readings are too low due to improper pump corrections of the sonde data, similar to the results of the BM-soundings obtained at Hohenpeissenberg.

Comparison of Japanese KC79 sondes with lidar at Tsukuba (see upper right panel of Figure 2.47 (Fujimoto et al., 1997)) show that the sondes tend to underestimate ozone between 18 and 23 km altitude, but overestimate ozone by about 2 to 4% in the region between 23-30 km and more than 5% above 30 km.

 

Figure 2.47 Average differences between non ECC sondes (Brewer Mast and KC79) and lidar or microwave radiometer as a function of altitude at a ± 24 hours coincidence criterion. The error bars are ± two sigma standard error.

 

2.5.6. Summary and Conclusions

In general all intercomparison studies have indicated that in the lower to middle stratosphere between the tropopause and ~ 28 km the three different sonde types show consistent results provided the individual measured sonde profiles have been normalised to ground based total ozone column measurements at the launch site. In this altitude range the precision of the various sonde types is within ± 3% , while any systematic bias compared to other ozone sensing techniques are smaller than ± 5%. For altitudes above 28 km the results are not so conclusive and the measurement behaviour of the sondes are different and cannot be generally characterised. The Brewer Mast sonde used by the established (long term record) stations (Hohenpeissenberg, Payerne, Uccle) show systematic under-estimations of ozone which increase with altitude (-15% at 30 km). For the ECC-sondes, there is some evidence suggesting that measurements agree with each other and to reference techniques to within ± 5%. The Japanese KC68/79 tends to overestimate ozone above 30 km. The data quality of sondes above 28 km is strongly influenced by the performance of the air sampling pump and its decaying efficiency at lower pressures. However, most of the intercomparison studies show that the performance of the ECC sondes between 28 and 35 km is still good and tends even to overestimate the ozone compared to lidar measurements. There are some experimental indications that for ECC-sondes there is a compensating effect due to evaporation of the KI-sensing solution which will cause an increase in concentration and may result in higher measured ozone. However, in general the sonde data above about 28 km are less reliable and should be used with caution for long-term trend evaluations, at least for the non-ECC-sonde types.

For the troposphere where the ozone concentrations are much smaller the results are not consistent at all. It is noted that validation of data from different types of sondes used in this study is only based on a very sparse number of experimental comparisons. Also, because ozone values are much lower in the troposphere than in the stratosphere, the impact of instrumental errors and instrumental variability is greater. Intercomparison campaigns done between 1970 and 1990 have shown systematic differences between sonde types typically varying from 10 to 15%. Recent campaigns (WMO-III, JOSIE, etc.) have shown that the Brewer Mast and KC79-sondes are less precise than the ECC-types and that the ECC-sondes are much more consistent than the other two types of sondes. The precision of the ECC-sonde is better than ± (5-10)% and shows a small positive bias of about 3%. The Brewer Mast and KC79 sonde showed precisions in the range of ± (10-20)% , but there are no indications of any bias larger than ± 5%.

There is some experimental evidence suggesting that the total ozone normalisation should not be used for ECC-sonde tropospheric ozone profile measurements while for the other sonde types it is less clear. Because this issue is still in an exploratory stage it is recommended that the normalisation be applied for any tropospheric data used in this trend assessment report.

Recent intercomparisons have shown that the observed differences between different sounding stations using the same sonde type are for a major part due to the differences in the preparation and correction procedures applied at the different launch sites. Fortunately some of the instrumental factors with the potential to influence the observed ozone trends regard post-flight data processing, and the data may be re-evaluated when the influences of these instrumental factors and their uncertainties are better understood. Although much progress have been made to improve the quality and homogeneity of the ozonesonde data since the last WMO Scientific Assessment of Stratospheric Ozone : 1994, there is still an urgent need to investigate and intercompare the instrumental performance of the different sonde types as well as to revise and achieve agreement on procedures for preparation and data processing.

 

Back to SPARC Ozone Report Homepage