GPS Atmospheric Water Vapor Measurements Without the Use of Local Barometers

Katherine J. Quinn and Thomas A. Herring
Dept of Earth, Atmospheric, and Planetary Sciences
Massachsetts Institute of Technology
Cambridge, MA 02139
Contact: katyq@chandler.mit.edu
Fall 1996 AGU Poster Presentation: G12A-06

Abstract

Field experiments have shown that GPS signals can be used to accurately measure vertically integrated atmospheric water vapor, or precipitable water (PW). Accurate surface pressure values are required at each GPS station where the water vapor is to be measured. Total zenith delay is estimated during GPS signal processing, this can be separated into zenith hydrostatic delay (ZHD) and zenith wet delay (ZWD). ZHD is proportional to the surface pressure, 10 mbar corresponds to approximately 23 mm of delay. ZWD is approximately proportional to the total water vapor, 10 mm PW corresponds to approximately 65 mm of ZWD. The zenith wet delay is found by subtracting the hydrostatic delay from the total delay, this is where accurate surface pressure measurements are required. 1 mbar of error in surface pressure could result in0.4mm error in precipitable water.
Accurate barometers are expensive pieces of equipment and very few of the GPS global tracking stations are equipped with them. However, the National Center for Environmental Prediction (NCEP) produces global meteorological data that may be used to estimate surface pressure, with no additional investment in equipment. We will examine the possibility of not using surface barometers by comparison of our NCEP derived values with existing barometer data at various International Geodetic Service (IGS) tracking stations.

Method

Interpolate upper level atmosphere fields of NCEP's aviation run of the global spectral model to find values at the IGS tracking station location.
Use a pressure model and the interpolated fields to calculate a pressure value at the IGS tracking station height above MSL.
Compare the interpolated station pressure to measurements made with a local barometer.
An error in surface pressure will map into an error in precipitable water estimate by a factor of 0.4 mm/mbar.
Assume that local barometers give accurate surface pressure to estimate precipitable water error due to using interpolated NCEP fields to calculate surface pressure.

IGS Tracking Stations

Location	Abbrev.	Latitude (N)	Longitude (E)	Height w.r.t. MSL (m)
Manama, Bahrain	BAHR	26.2091	50.6081	10.1674
Fairbanks, AK	FAIR	64.9780	212.5010	307.2176
Greenbelt, MD	GODE	39.0217	283.1732	49.3216
Kitab, Uzbekistan	KIT3	39.1348	66.8854	659.5424
Kokee Park, HI	KOKB	22.1263	200.3350	1158.0989
Potsdam, Germany	POTS	52.3793	13.0661	104.6891
Koetzting, Germany	WETT	49.1442	12.8789	620.1434

Figure 1: World map of locations of IGS tracking stations

Results

Station	Surface Pressure (mbar)		Precipitable Water (mm)
Station	RMSE	Offset	RMSE	Offset
BAHR	6.2452	2.2992	2.2099	0.8136
FAIR	7.8233	2.4519	2.7683	0.8676
GODE	4.7760	2.8065	1.6900	0.9931
KIT3	2.6257	-1.3219	0.9291	-0.4678
KOKB	0.6632	-0.3175	0.2347	-0.1124
POTS	1.8210	0.2445	0.6444	0.0865
WETT	2.0970	1.1673	0.7420	0.4131

The following figures have a top plot and a bottom plot. The top plot is a comparison of surface pressure derived from interpolated NCEP atmospheric fields and surface pressure locally measured with barometers at an IGS tracking station from 6/1/96 to 11/1/96. The bottom plot is a residual plot of the comparison. Left hand scale is surface pressure error (mbar) and right hand scale is estimated precipicatable water error (mm).

Figure 2: WETT
- Figure 2a: WETT, 8/15/96 - 9/15/96. Example of good fit between derived surface pressure and measured surface pressure.
- Figure 2b: WETT, 6/1/96 - 7/1/96. Example of not so good fit between derived surface pressure and measured surface pressure.
Figure 3: GODE
- Figure 3a: GODE, 7/1/96 - 8/1/96. This shows a section where the local barometer had problems.
Figure 4: KOKB
- Figure 4a: KOKB, 8/15/96 - 9/15/96. Note the diurnal cycle for stations in the tropics. At one point the local barometer doesn't appear to keep up with the high frequency pressure fluctuations.
Figure 5: POTS
Figure 6: KIT3
Figure 7: FAIR
Figure 8: BAHR

Conclusions

To provide a real contribution to the meteorological community we would want to have the error in the estimate of precipitable water of order 1mm. Comparisons of surface pressure derived from NCEP global atmospheric fields to surface pressure measured by local barometers are very favorable at a number of the IGS stations. Problems exist for stations at higher latitudes (FAIR) where the NCEP models are known to be less accurate. It can also be seen at some of the stations (BAHR, GODE) that assuming the local barometers are accurate can be misleading, in-situ measurements may have considerable errors.
We envision that GPS estimates of total zenith delay could be fed directly into global atmospheric operational models without pre-processing into precipitable water. The atmospheric models would have algorithms incorporated into them to assimilate the total zenith delay using the atmospheric fields in the model to convert zenith delay into precipitable water.