A
comparison of the 2005 total ozone measurements done with |
|
version 1.0 |
Summary | The year 2005 total ozone measurements made by 4 different instruments, two ground located and two satellite-born, gives statistical meaningful correlation factors mostly exceeding 0.90. A best multiplier of 0.95 is suggested to apply to the meteoLCD Microtops # 5375 readings if the Brewer measurements published as Uccle Woudc mean readings are to be taken as a reference source. The differences between the satellite and Brewer instruments are of a similar magnitude than those between Microtops and Brewer. Only data from January to May 2005 were available from Sciamachy. |
1. Reminder on Uccle and Diekirch locations and satellite characteristics:
Please read here
for the relevant details. The monthly Sciamachy data files [2] cover the whole
globe; only data from Jan to May 2005 were freely available at the time of writing this
report (May 2006);
NASA has issued a warning [5] regarding the use of Toms data for longtime drift
analysis. The Toms datafile [1] is complete, with 365 measurements corresponding
to a daily Uccle overflight. Both Uccle
and Microtops datafiles have gaps that are often non-overlapping (see table1).
At Diekirch a very mall subset of the measurements were done from end of July to
end of August 2005 with the older #3012 model and a different operator; an
overlap region with both instruments used allowed to apply a corrective
multiplier to make these measurements compatible with model #5375
characteristics.
The conclusion from the years 1997 to 2002 comparison between Microtops #3012 and the supposed correct Brewer DS measurement were to apply multiplier of 0.93 - 0.95 to the Microtops readings to give a good agreement with the Brewer; the 2003 report suggested no need for a multiplier for Microtops #5375 (who started operation in May 2003, and has not yet been recalibrated by the manufacturer).
2. Total O3 readings from the 4 instruments:means and standard deviations
Table 1 shows the number of records available, means, standard-deviations and the mean of the sum of the squared differences with Microtops:
Table 1 |
January to May 2005 |
January to December 2005 |
||||
Instrument | number of records | mean +/- std | mean of sum of squared differences (#) | number of records | mean +/- std | mean of sum of squared differences (#) |
Toms | 151 | 332.5 +/- 48.7 | 1060 | 365 | 313.0 +/- 44.3 | 1252 |
Sciamachy | 60 | 346.7 +/- 41.3 | 738 | NA | NA | NA |
Uccle_Woudc | 237 | 349.1 +/- 40.3 | 233 | 237 | 324.9 +/- 39.4 | 426 |
Uccle_MkII | 52 | 336.3 +/- 46.2 | 649 | 181 | 316.2 +/- 39.2 | 898 |
meteoLCD | 70 | 351.0 +/- 47.1 | 137 | 334.8 +/- 40.6 |
(#) differences versus meteoLCD
Table 2 shows the effect of a two restrictions of the comparison: Uccle_Woudc_keep is restricted to the readings of Uccle and Diekirch done at times less than one hour apart (these readings will have the "keep" suffix), and Uccle_MKII_keep_DS to the similar MarkII readings, with the supplementary condition of a direct sun pointing (DS):
Table 2 |
January to May |
January to December |
||||
Instrument | number of records | mean +/- std | mean of sum of squared differences (#) | number of records | mean +/- std | mean of sum of squared differences (#) |
Uccle_Woudc_keep | 30 | 337.4 +/-44.1 | 218 | 81 | 313.8 +/- 38.9 | 431 |
Uccle_MkII_keep_DS | 07 | 318.6 +/- 30.3 | 133 | 28 | 301.6 +/- 33.0 | 708 |
meteoLCD_keep | 30 | 348.6 +/-45.0 | 81 | 331.4 +/- 38.4 |
(#) differences versus meteoLCD
Conclusion: meteoLCD Microtops readings are closest to the mean Uccle_Woudc data; the cumulative deviation from the direct sun means are noticeably higher, but the relatively small number of available records (28) makes usage of these questionable. The increase of the mean of sum of squared differences between the January-May and whole year data points to an increasing drift of the Microtops during 2005, assuming the Brewer remaining stable (what has been confirmed by Dr. De Backer, personal communication). The Uccle_Woudc readings are the official data published by Uccle; it seems reasonable to focus the comparison between Diekirch and Uccle on these data, and not so much on the (unofficial) MKII readings, which deviate markedly from the former.
3. A possible drift between Microtops during 2005
Conclusion: All mean differences between Microtops and the other sensors are positive, as are the linear trends over 2005. Probably Microtops had a small positive 12 to14 DU drift (depending on the reference source used) during 2005.
4. Correlations and linear calibration factors
Table 3 gives the linear correlation factors, which are all significant at the 5% level:
Table 3 | Sciamachy | UccleWoudc | UccleWoudckeep | UccleMKII | UccleMKII_DS | Microtops |
Toms | 0.90 | 0.95 | 0.96 | 0.90 | 0.94 | 0.94 |
Sciamachy | 0.93 | 0.97 | 0.78 | 0.95 | ||
UccleWoudc | 0.93 | 0.98 | 0.95 | |||
UccleWoudckeep | 0.96 | 0.97 | 0.96 | |||
UccleMKII | 0.88 | |||||
UccleMKII_DS | 0.93 |
One should note the poorer correlations with UccleMKII; the author has no explanation on this for the moment.
Linear regressions forced through the origin will be computed to give simple calibration multiplicators. Table 4 holds the linear regression slopes for all sensors and subsets (x- values = Microtops, y-values = other sensor); the Goodness of the Fit (GOF) is given between parenthesis
Table 4 | Sciamachy | UccleWoudc | UccleWoudckeep | UccleMKII | UccleMKII_DS | Microtops |
Toms | 0.98 | 0.96 (0.94) | 0.95 (0.96) | 0.97 (0.90) | 0.95 (0.94) | 0.91 (0.93) |
Sciamachy | 0.98 (0.93) | 0.98 (0.95) | 1.02 (0.73) | NA | 0.94 (0.93) | |
UccleWoudc | 1.02 (0.93) | 1.01 (0.97) | 0.95 (0.95) | |||
UccleWoudckeep | 0.95 (0.96) | |||||
UccleMKII | 0.93 (0.86) | |||||
UccleMKII_DS | 0.93 (0.93) |
These slopes represent the multiplier to apply to Microtops data to obtain calibrated data, the reference sensor assumed to be correct; a multiplier of 1 would imply a perfectly calibrated instrument. The smaller multiplier for the UccleMKII comes as a surprise, and the restriction to the direct-sun readings does not improve the multiplier (but increases the goodness of the fit). As there are only 28 data to compare UccleMKII_DS and Microtops, the most sensible conclusion would be to to accept the UccleWoudckeep readings as a reference, and to apply a multiplier of 0.96 to the 2005 Microtops results. It should also be noted that the multiplier and GOF for Microtops/Toms and UccleWoudckeep/Toms are the same; Sciamachy/Ucclewoudckeep is higher than Microtops/Ucclewoudckeep, but the GOF's are close.
The next figures 6,7 and 8 show the graphs with the linear regressions
5 Conclusions The Microtops II #5375 measured mean total ozone column is best compared to the Uccle mean total ozone measurements as published at the WOUDC, retaining only data whose measurement times differ less than 1 hour with Microtops. The Microtops sun photometer probably suffered from of a small positive drift during 2005. This report suggests to apply a multiplier of 0.95 to give the best correspondence with the Uccle mean Woudc data. Different from previous years, the Uccle_MKII (direct sun or not) readings give a poorer fit (with a suggested multiplier of 0.93), as do the measurements from the satellite born Toms and Sciamachy instruments. Differences between Microtops and the satellites are however of a similar magnitude as those between Uccle and the satellites. |
Special thanks to Dr. De Backer of the Royal Meteorological Institute of Belgium (RMI) for providing the Uccle Brewer data and his continuing help.
1. | TOM's data are available at: ftp://jwocky.gsfc.nasa.gov/pub/eptoms/data/overpass/OVP053_ept.txt |
2. | Sciamachy monthly data files (whole globe): http://www.temis.nl/protocols/o3col/o3month2005.html |
3. | Francis Massen : Influence of intentionally mis-pointing on Microtops II readings [2004] |
4. | Dr. Hugo De Backer, RMI: personal communication |
5. | NASA Toms calibration problems: http://toms.gsfc.nasa.gov/news/news.html#nov18 |
6, | The Uccle (station53) 2005 WOUDC data. |
7. | The meteoLCD (station 412) 2005 WOUDC data. |
8. | RMI website with Brewer total O3 graph |
All calculations and graphs have been made using Statistica 7.1 by Statsoft.
History of versions:
1.0 original version 02 May 2006
Please send all mail to francis.massen@education.lu |