1. Comparability and Reproducibility of RO Data
Shu-peng Ho1,2, Ying-Hwa Kuo1,2, UCAR COSMIC team, Jens Wickert, GFZ team,
Gottfried Kirchengast, Wegener. C., Chi Ao, Tony Mannucci, JPL teams,
Cheng-Zhi Zou3, and Mitch Goldberg3
1. National Center for Atmospheric Research,
2. University Corporation for Atmospheric Research/COSMIC
3. NOAA/NESDIS/Center for Satellite Applications and Research

2. Acknowledgements
J. Wickert from GFZ, Gottfried Kirchengast from Wegener Center, Chi Ao, Tony Mannucci from JPL team are acknowledged for continuous collaboration
Long term processing and maintaining CHAMP RO data from GFZ are especially acknowledged
CHAMP, our „good old working horse“ in space since 2000 2

3. What are the uncertainties for using GPS RO data for climate
Motivation:
What are the uncertainties for using GPS RO data for climate
monitoring ? Can we use GPS RO data to inter-calibrate other climate
data ?
GPS RO data for climate monitoring:
Raw observation is SI traceable, high vertical
resolution, insensitive to clouds and
precipitation
Good temporal and spatial coverage
Comparability:
High precision
Long term stability
c) Reproducibility: Reasonable uncertainty
among data processed from different centers
2. Outlines :
Challenges to define/validate a global trend : a) Satellite b) Radiosonde
Characteristics of COSMIC GPS RO data for climate monitoring
Compare refractivities generated from different centers
3. Conclusions and Future Work
Slide 3
Shu-peng Ben Ho, UCAR/COSMIC

4. Challenges for defining Climate Trend using satellite data
Satellites: Comparability and Reproducibility ?
1) Not designed for climate monitoring
2) Changing platforms and instruments
(No Comparability)
3) Different processing/merging method
lead to different trends Due to the differing methods used to account for errors before merging the time series of eleven AMSU/MSU satellites into a single, homogeneous time series, these derived trends are different from different groups (RSS vs. UAH).
(No Reproducibility)
Slide 4
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC
Shu-peng Ben Ho, UCAR/COSMIC

5. Challenges for defining Climate Trend using Radiosonde Data
We need measurements with high precision, high accuracy, long term stability, reasonably good temporal and spatial coverage as climate benchmark observations.
Radiosonde: Comparability and Reproducibility ?
1) Measurements will be affect on instrument type, location, and
the environments (No Comparability)
2) changing instruments and observation practices
3) Uneven spatial coverage, limited spatial coverage especially over the oceans;
4) Uneven temporal coverage
Slide 5

6. Comparability of COSMIC data from different receivers
Challenges:
a. Extreme on-orbit environment
b. Same atmospheric path
c. Temporal/spatial mismatch
d. Reasonable sample number
e. Temporal/spatial dependent
biases
Using FM3-FM4 pairs in early mission
Need to quantify all
COSMIC-COSMIC pairs
Within 25 km
(Ho et al. TAO, 2008)
Dry temperature difference between FM3-FM4 receivers
Slide 6
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

7. Global mean FM#3-FM#4 dry temperature difference (K)
Mean dry temperature difference between FM3-FM4 is < 0.1 K
Not location dependent
Not time dependent
Natural variability and sampling errors dominate MAD
Latitude
Latitude
Latitude
Dry Temperature
Difference (K)
Median Absolute
Deviation (K)
Sample Numbers
Slide 7

8. Comparability of CHAMP to COSMIC:Long-term stability
Global COSMIC-CHAMP Comparison from
Comparison of
measurements between old
and new instrument
CHAMP launched in 2001
COSMIC launched 2006
Challenges:
a. Different inclination angle
b. Different atmospheric
paths
c. Temporal/spatial mismatch
d. Reasonable sample number
Need to have stable
calibration reference
Within 90 Mins
and 250 Km
Within 90 Mins
and 100 Km
Within 60 Mins
and 50 Km
Slide 8

9. Approach d(tau)/d(lnP) Approaches:
1. Apply CHAMP and COSMIC soundings to AMSU forward model to simulate AMSU TLS
2. Match simulated GPS RO TLS to NOAA AMSU TLS to find calibration coefficients for different NOAA satellites so that we can
d(tau)/d(lnP)
Slide 9
Shu-peng Ben Ho, UCAR/COSMIC

10. NOAA 18 AMSU Ch9 Brightness Temperature
The mean weighted dry temperature difference between COSMIC and CHAMP from 300 mb to 10 mb is about 0.07 K b c
NOAA 18 AMSU Ch9 Brightness Temperature
Slide 10

11. Monthly Mean Climatology
Reproducibility of GPS RO data 
Similar measurement errors:
Thermal noise
Ionospheric calibration
Different:
Orbital errors
Initial Integral of Abel
Inversion algorithm
(from bending angle to refractivity)
Monthly Mean Climatology
Quality control method
Slide 11

12. Fractional refractivity from JPL, UCAR and GFZ
Monthly, 5 deg-lat, 200-meter mean refractivity profiles from
Fractional refractivity from JPL, UCAR and GFZ
Bias=-0.05%
Std = 0.45%
Bias=0.001%
Std = 0.45%
100x
100x
Bias and std from 30 km to 8 km
Slide 12
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

13. 8-30 km Global North pole North H. Mid-lat Tropics South H. Mid-lat
South pole
Slide 13

14. 20-30 km Global North pole North H. Mid-lat Tropics South H. Mid-lat
South pole
Slide 13
Slide 14
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

15. 20-30 km Global North pole Tropics North H. Mid-lat South pole
South H. Mid-lat
Slide 19
Slide 15
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

16. 8-30 km Global North pole Tropics North H. Mid-lat South pole
South H. Mid-lat
Slide 16
Fig. 1

17. The uncertainty of the trend of fractional N anomalies is within
+/-0.04 %/5 yrs.
What is the cause of small trend difference ?
Slide 17
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

18. 20-30 km Global North pole Tropics North H. Mid-lat South H. Mid-lat
-0.68%
-0.70%
-0.43%
South pole
Slide 19
Slide 18
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

19. 20-30 km Global North pole North H. Mid-lat Tropics South H. Mid-lat
South pole
Slide 13
Slide 19
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

20. Bias and MAD from 30km to 8 km
GFZ and UCAR pixel to pixel refractivity comparison from
Mean MAD from 8 km to 30 km = 0.16%
Bias and MAD from 30km to 8 km
Slide 20

21. 8-30 km Global North pole Tropics North H. Mid-lat South H. Mid-lat
South pole
Slide 21

22. 8-30 km Global North pole North H. Mid-lat Tropics South H. Mid-lat
South pole
Slide 22

23. Conclusions and Future Work
It is a great challenge to use current available datasets to construct reliable climate records.
The less than 0.1 K precision of COSMIC will be very useful to inter-calibrate AMSU/MSU data.
The long term stability of GPS RO data is very useful for climate monitoring.
Although different centers using different inversion procedures and initial conditions to derive refractivity, and using the different quality control criteria to bin the datasets, the mean bias for JPL-UCAR pairs is -0.05%, and for GFZ-UCAR pairs is 0.001%.
The uncertainty of the trend of the fractional N anomalies is within +/-0.04 %/5 yrs (+/-0.06 K/5 yrs). And the major causes of uncertainties between these trends are from sample profiles used by different centers.
GPS RO temperature shall be very useful to calibrate measurements from other
satellites.
Slide 23
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC

24. Conclusions and Future Work
Can we use the NOAA satellite measurements
calibrated by GPS RO data to calibrate
multi-year AMSU/MSU data ?
(Ho et al. GRL, 2007)
Slide 24
Copy right © UCAR, all rights reserved
Shu-peng Ben Ho, UCAR/COSMIC