1. A global model for the conversion of ZWD to IWV
Rózsa, Sz.; Juni I
Department of Geodesy and Surveying, Budapest University of Technology and Economics

2. Outline Introduction of the models studied; Methodology;
Global Models for the Determination of the scale factor ZWD/IWV;
Validation with RS observations
Conclusion and Outlook

3. Conversion of ZWD to IWV
ZTD is estimated in GNSS processing;
ZHD is modeled based on air pressure data (observation/interpolation/model)
ZWD = ZTD – ZHD
IWV = ZWD × Q

4. Conversion of ZWD to IWV
Usually two approaches (others exist):
Bevis et al (1992) expresses the Tm as a linear regression of Ts
(North-american RS data)
Emardson-Derks (2000) expresses the scale factor as a polynomial function of Ts:
(European RS data)

5. Conversion of ZWD to IWV
Similar models exist for different continents, regions
These models:
usually neglect climatic effects within the studied region
are not available globally (places with lack of RS observations)
A seamless global model could assist the GNSS based IWV estimation and the validation of the results.

6. Methodology
Global grids of ECMWF ERA-Interim data for the period of
monthly mean solutions (120 data sets)
37 pressure levels
Temperature, relative humidity and geopotential
1°×1° resolution
Numeric integration to compute
Vertical interpolation;
ZWD;
IWV;
Tm (temperature weighted by the water vapour density);

7. The global TmTs model
Approach 1: Estimate the parameters of the TmTs linear regression for each grid point

8. The global TmTs model
The relative difference between the original Bevis et al. and the TmTs model for January, 2001:
mean:

9. The global polynomial model
Approach 2: Estimate the model parameters of polynomial Q=f(Ts) with LSA for each grid point.

10. The global polynomial model

11. The global polynomial model
The relative difference between the original Emardson-Derks polynomial and the global polynomial model for January, 2001:
mean:
Model artifacts (10%, -10%)

12. Validation with RS comparisons
Altogether 20 globally distributed RS stations
period – independent from the model data used for the derivation
ZWD, IWV computed by numerical integration
QRS=IWV / ZWD
Qmodel=f(TS) – where TS is the surface temperature stemming from RS profiles
QRS-Qmodel residuals computed

13. The RS network
RS sites should be available in the NOAA RS database with small gaps
tried to have a homogeneous coverage, but still large gaps occur
Altogether 58,690 RS profiles were used

14. Results
Hong Kong

15. Results
Curitiba

16. Results
Budapest (HUN)

17. Results Relative mean bias of Q values (wrt RS observations) Model
Bevis
0,7%
TmTs
1,3%
Emardson-Derks
0,9%
Polynomial
0,4%

18. Conclusion and outlooks
Two global models have been derived for the estimation of the scale factor as a function of Ts;
Can be used when no RS observations are available in the region for the local fitting of the Q formulae;
The global polynomial model provided the smallest global mean bias (0.4%);
The improvement was detected mainly in the tropical region;
There are some local artifacts mainly in South America, which still need to be investigated;
Data sets and software available soon at: for testing.

19. Thank You for Your Attention
Szabolcs Rózsa, Ildikó Juni
Department of Geodesy and Surveying
Budapest University of Technology and Economics
H-1111 Budapest, Muegyetem rkp. 3
URL:
Software and models will be available at