Corrections for Bias of Chinese Upper-air Measurements
Guo Yatian, Huang Bingxun
Chinese Academy of meteorological Sciences
Hu Deyun, Fan Zhende
China Meteorological Administration, Beijing, China
1 Introduction
China is making a sustained effort in improving the accuracy of the network measurements and cost-effectiveness of radiosonde. A new generation of radiosonde has been deployed at Beijing station in January 2002. The replacement plan of radiosonde and ground equipment of upper-air network will be implemented in the next 3 years. In the meantime,thesoftware of the mechanical radiosonde (GZZ2), which is still used as a lead, has been updated significantly. The necessary corrections for various errors have been introduced or revised to reduce the bias of upper-air measurements. The updated software has been installedat 50 stations since July 2000 and extended to remainder stations since January 2001.
The software revision is based onlong time investigations to various errors of the GZZ2 sonde. The extensive investigations demonstrate that the factors affecting the bias of GZZ2 measurements can be classified into objective and subjectivetypes.
The objective factors:
The mechanical radiosonde had been eliminated in the world network except in China.
WMO uses the First Guess (FG) field of ECMWFinstead of the analytical field (AN) as the reference to monitor the quality of network. The biases from the FG are bigger than that from the AN.
The subjective factors:
GZZ2 as a mechanical sonde has more radiation and lag error than sonde using electric elements.
Due to much manual operation in the data reducing, only anaverage radiation error has been corrected while the lag error has been left uncorrected.
The cuprous chloride and magnesium battery placed in the up part of the sonde box has some thermal impact on the temperature element.
The suspension rope of the GZZ2 is only 12m long. Therefore the sonde is also suffered from the thermal effect of the balloon.
The aneroid capsule used by GZZ2 has appreciable temperature coefficient. The releasing comparisons between GZZ2 and otherhigh quality radiosondes had revealed that GZZ2 indicated lower pressure than others.
The revision of GZZ2 software is based exactly on above-mentioned investigation results. The main contents of revision involve amending the gn from 9.80 to 9.80665ms-2, improving radiation correction method, correcting the drift of pressure calibration, correcting the lag error of the temperature unit and correcting the thermal effect of the sonde box and balloon to the temperature element.
A bilateral flight comparison was conducted in Zhengzhou of China between GZZ2 and Vaisala’s RS80 sondes in March 1996. In order to remain the error characters of GZZ2, the two sondes were suspended separatelybelow two balloons and released simultaneouslywith a distance no more than 50m. The comparison results showed that there were significant differences of geopotential height at standard pressure levels between GZZ2 and RS80 and the differences were close to the WMO monitoring results of (OB-FG) bias of GZZ2 for 1996 (Oakley 1998).
To verify and optimize the correction model of the new software, the correcting values were compared with comparison flight dataat Zhengzhou and withthe monthly (OB-FG) biases of 13 stations scattered in China for 1996 obtained from T. Oakley.
The following paragraphs illustrate the comparison resultsbetween the correcting valuesand bilateralflight data at Zhengzhou and that between the correcting values and the monthly (OB-FG) biases of 13 stations for 1996. The effectiveness of the software revision will beverified through a extensive discussion about the (OB-FG)biasvariations of GZZ2 before and after correction, and the bias variations of other radiosondes widely used in the world.
2Comparison between corrections and bilateral flight data at Zhengzhou
The thick line in figure-2.1 shows geopotential differences at standard pressure levels between GZZ2 and RS80 at 00Z and 12Z from bilateral comparison at Zhengzhou. The thin lines separated by the thick line show the reliability limits of the geopotential difference by using the standard deviation divided by the square root of the number of statistics. The symbol circles are correction values of the new software to the old one. It is shown that the geopontential height of GZZ2 at 100hPa is 42m higher than that of RS80 in darkness (12Z) and 70m higher in daytime (00Z). It is also demonstrated that almost all of the calculated correction values fall in the reliability limits. That is to say the correction values are very close to the differences of actual flight comparison between GZZ2 and RS80.
Figure 2.1 Comparison between corrections of new software and results of Zhengzhou bi-comparison
3 Comparison between corrections and (OB-FG) bias for 1996
Figures-3.1 to 3.3 show the variations of geopontential biases of 13 stations at 100hPa with longitude, months and latitude respectively for 1996 before and after correction applied.
Figure 3.1 Bias variation with longitude. Figure 3.2 Bias variation with months. Figure 3.3 Bias variation with latitude.
It is clear that the geopontential biases of 13 stations at 100hPa after correction have been reduced significantly and are close to the ‘zero line’ and obviously unvaried with longitude, latitude and months.
4 Verification of the effectiveness of the software revision
How is the situation of bias of Chinese network after the correctionbeing applied? Following demonstrates extensive analysis from various angles to the (OB-FG) statistics from 1988 to 2001 by WMO.
4.1Variation of (OB-FG) biases of GZZ2 before and after software revision
Figure 4.1 shows the variation of OB-FG biases of GZZ2 at 100hPa before and after correction. The data are from statistics of recent WMO Rapporteur on Radiosondes Compatibility Monitoring, J. Elms.
Figure 4.1 Bias variation of GZZ2 Figure 4.2 Bias variation of GZZ2 and RS80
It is seen that the biases for 50 corrected stations have been reduced 60 meters at third quarter and 40 meters at fourth quarter of 2000 respectively and have achieved the desired results. In addition, the bias variation with seasonafter updating of the software since January 2001 is less obvious than before. However, the average (OB-FG) bias of GZZ2 is about –20m and not near the ‘zero line’. Whether the Chinese measurements have been over-corrected or not? In order to finding out the truth, we have reviewed the (OB-FG) biases statistics of some widely used radiosonde from different angles in the next paragraphs.
4.2Biases of RS80 used in Hongkong and Taibei of China and Mongolia
Figure 4.2 shows variations of biases of RS80 used in Hongkong and Taipei of China and in Mongolia.It can be seen that the biases of RS80 at 100 hPa are also near–20msince 2000.
4.3Variation of biases of some main radiosonde types
In order to check the stability of FG field, we have reviewed the (OB-FG) biases of some main types of radiosonde used all over the world according to the (OB-FG) statistics of WMO since 1988.
Figure 4.3 shows the variation of (OB-FG) of RS80 (partly RS90) of Vaisala used in some areas.
Figure 4.3 Variation of (OB-FG) of GZZ2 and RS80 at 100hPa(00Z)
It is shown that only the (OB-FG) of RS80 used in West Europe has a bias of±5m. In other areas, such as Canada, USA (including Alaska), Australia, Malaysia, Vietnam, Singapore, Mongolia,HongKong and Taipei of China, there are obvious and consistent variations of (OB-FG) of RS80 since 1988. Considering the latest 2 years, the biases of RS80 used in various areas are within the range of 0 to –20m, and the bias of GZZ2 is in the lower limit of the range.
Figure 4.4 shows the variation of biases of VIZ type sonde used in USA and other areas and Japanese sonde. It is seen that VIZ type sonde has significant seasonal variation similar to GZZ2. And the average bias in native land of USA has descended 10m since 1997. The bias variation of Japanese sonde is similar to RS80 and GZZ2. However, the (OB-FG) in southern is 10-20m lower than that in northern Japan. The biases of VIZ used in Thailand or VIZ type sonde used in Korea fluctuate significantly within a wide range. Considering the latest 2 years, the (OB-FG) bias of Japanese sonde is also within the range of 0 to –20m.
Figure 4.4 Variation of (OB-FG) of VIZ type sonde at 100hPa(00Z)
4.4Variation of (FG-REF)
As a results of the 84/85 WMO comparison experiments, a geopotential ‘reference measurements’ was defined as the average of the geopotential measurements from the VIZ and Vaisala RS80 radiosonde during darkness. This standard is referred as the WMO reference (REF) in the WMO statistics. However, the WMO comparisons were conducted at limited areas and times. Therefore, now WMO uses FG of ECWMF as reference in the statistics. Unfortunately, there are also errors in the FG field and the errors are varied with areas and times. In the WMO Radiosonde Compatibility reports in 1989 (kitchen) and 1993 (Oakley), the stability of (FG-REF) had been ‘calibrated’ by use of the (OB-FG) biases of some widely used radiosonde types. In the 1989 report, the (FG-REF) value at 100hPa was 25m. But in the 1993 report, the (FG-REF) value was down to 5-10m. This may be used to explain the upward variation of (OB-FG) values in the period. In the 1998 report (Oakley), the (FG-REF) value was assumed within 5-10m according to the stable (OB-FG) values of some widely used radiosonde types. Moreover, it was indicated by the Rapporteur that a change of at least 10m in the overall systematic error in FG field should be considered. In the 1998 report, an exception was indicated that the average (OB-FG) bias values for the fourth quarter of 1997 were more than 10m lower than that for the previous years. Since 1995, more and more VIZ sonde were breplaced by Vaisala RS80 in the North American network. The REF was defined as the average of the geopotential measurements from the VIZ and Vaisala RS80 radiosonde during darkness as indicated before. Therefore the (OB-FG) bias average of VIZ and RS80 used in USA perhaps reflects the variation of FG field to some extent. From figure 4.5, it can be seen that the average bias of RS80 and VIZ in 1995-1996 period was close to ‘zero line’, but descended to –10m after 1997
Figure 4.5Variation of (FG-REF) and the average bias of VIZ and VRS80
4.6Summary of verification of correction
In summary, the (OB-FG) biases of world–wide radiosondes from 1988 have showed a significant change only with an exception in West Europe. There was a high peak in 1995-1996 period, but the biases dropped to the lowest in 1997-1998. It is obvious that this variation could not be attributed fully to the changes of radiosonde or observation method and the FG field could not be assumed have zero systematic error. However, considering the latest 2 years, the OB-FG biases of RS80, VIZ and Japanese sonde are within the range of 0 to –20m. And the OB-FG bias of GZZ2 after correction is about –20m. So as a primary judgment, the GZZ2 is assumed over-corrected about 10m. More monitoring results are required before a reliable conclusion can be made.
5Conclusion
According to the WMO monitoring results of the (OB-FG) bias of GZZ2 before and after revision of software, it is obvious that the software revision is effective to reduce the systematic bias of GZZ2. After revision, the bias of GZZ2 is close to the biases of Vaisala RS80 and VIZ of US. In addition, the seasonal variation of the bias of GZZ2 has been reduced notably.
To monitoring the quality of worldwide network just using FG field as reference seems not very ideal. Except the bias of RS80 used in West Europe keeps to be stable, the variation of biases in other areas is very significant and difficult to explain wholly by the changes of radiosonde or observation methods. Perhaps, a comprehensive monitoring method using FG, and AN (analytical field) and others as reference will give us a more ideal monitoring results.
6Acknowledgements
The authers wish to thank Mr T. Oakley and J.B.Elms of UK for their support in time with WMO monitoring results of Chinese network for 1996 and 2000-2001.
7Referances
Kitchen, M. : Compatibility of Radiosonde Geopontential Measurements 1989.
Observational Services Memorandum (OSM) No.38, Meteorological Office, Bracknell, 1989.
Oakley,T. : Compatibility of Radiosonde Geopontential Measurements 1990,1991 and 1992.
WMO Instruments and Observing Methods Report No.56, 1993.
Oakley,T. : Compatibility of Radiosonde Geopontential Measurements 1995,1996 and 1997.
WMO Instruments and Observing Methods Report No.56, 1998.