Water vapor imaging in the troposphere by combination of GNSS

Water vapor imaging in the troposphere by combination of GNSS
Water vapor imaging in the troposphere by combination of GNSS occultation data
and ground-based GPS measurements
Ulrich Foelsche and Gottfried Kirchengast
Institute for Geophysics, Astrophysics, and Meteorology, University of Graz (IGAM/UG), Universitaetsplatz 5, A-8010 Graz, Austria
phone: +43 316 380 8590, fax: +43 316 380 9825, email: ulrich.foelsche@kfunigraz.ac.at
Retrieval scheme
The potential of Global Positioning System (GPS) measurements for accurately estimating vertically
and slant-path integrated water vapor between GPS satellites and ground-based receivers has been
demonstrated recently (e.g., [1],[2],[3]).
Global Navigation Satellite System (GNSS) based radio occultation, on the other hand, has been
shown to deliver accurate near-vertical profiles of atmospheric variables such as temperature and
humidity with high vertical resolution (e.g., [4]). Height resolving imaging of atmospheric water
vapor becomes feasible when occultation profiles from spaceborne receivers in Low Earth Orbits
(LEO) are combined with ground-based GNSS data from a co-located receiver network.
We developed a two-dimensional, height-resolving tomographic imaging technique following the
Bayesian approach for optimal combination of information from different sources.
The reconstruction plane, defined by the occultation rays, is divided into picture elements (pixels)
with assumed constant water vapor density (see Fig. 1). For the ground measurements, the rays can
be considered as straight lines (the integrals in the forward problem degenerate into simple sums of
densities times ray lengths in each pixel), but for the occultation rays bending cannot be neglected.
The latter are thus not incorporated directly into the inversion, but as "a priori" information via
optimal estimation [5], exploiting that the occultation delivers a reliable mean refractivity profile.
The corresponding mean water vapor density profile (representative for the entire retrieval domain)
can be computed given additional temperature profile information (e.g., from the latest ECMWF
analysis). The accurately measured vertical integrated water vapor (IWV) can be used to adjust the
water vapor density profile to match in integral this IWV value (see Fig. 5d).
We show representative results, using simulated GNSS-based water vapor measurements from LEO
and ground, derived from simple synthetic refractivity fields (Figs. 3 and 4) and from a realistic
refractivity field based on a European Centre for Medium-range Weather Forecasts (ECMWF)
analysis (Fig. 5).
xretr = xap + SretrATSε-1(y - Axap)
Sretr = (ATSε-1A + Sap-1)-1
xretr = retrieved state vector
xap = a priori state vector (occultation-derived)
Sretr = retrieval covariance matrix
Sε = measurement covariance matrix
Sap = a priori covariance matrix
In order to basically investigate the
performance of the retrieval algorithm,
we directly used synthetic water vapor
fields and corresponding SIWV values.
The a priori field used in this case was
simply the mean density profile
extended over the entire reconstruction
domain. The latitude range used was
arbitrarily chosen centered at 45° (Figs.
3 and 4).
In the ECMWF case (Fig. 5) we used the
full forward model scheme outlined.
Fig. 3a: Synthetic exponential atmosphere with linear
horizontal gradient. The exponential decrease with
height is described by a climatological (constant) water
vapor scale height of 2 km.
Fig. 3b: Optimal estimation retrieval of the synthetic
exponential atmosphere with linear horizontal gradient,
assuming an rms SIWV error of 1.5 kg/m2. Retrieval
height-domain 0 - 10 km, 20 x 10 pixels.
Fig. 3c: Relative difference between optimal estimation retrieval (Fig. 3b) and the original water vapor
density field (Fig. 3a).
Fig. 4a: Model water vapor density for an isolated
Gaussian blob with a vertical half-width of 1 km, and a
horizontal half-width of 0.15° (~17 km).
Fig. 4b: Optimal estimation retrieval of the water vapor
density field shown in Fig. 4a, assuming an rms SIWV
error of 1.5 kg/m2.
Fig. 4c: Absolute difference between retrieved (Fig.
4b) and model water vapor density (Fig. 4a), respectively, for the scenario of an isolated Gaussian blob.
Fig. 5b: Optimal estimation retrieval of the water vapor
density field shown in Fig. 5a (Florida case). Height
range 0 - 6 km, 12 x 10 pixels (see Fig. 1).
Fig. 5c: Relative difference between retrieved (Fig. 5b)
and model water vapor density (Fig. 5a) - Florida case.
Fig. 5d: Averaged profile of the water vapor density
field shown in Fig. 5a (solid line), density profile
obtained by mimicked radio occultation (dotted), and
occultation profile after IWV adjustment (dashed), the
latter used as a priori profile.
Forward model scheme
Model atmosphere (e.g.
ECMWF analysis)
Surface pressure
Fig. 5a: Model water vapor density, derived from
ECMWF analysis data over Florida for October 20,
1995 (12 UT time layer, T213L31 resolution).
Hydrostatic equation
Optical path integral
Zenith hydrostatic delay
Path meas. rms error
Mapping function
Excess path delay
Slant hydrostatic delay
Slant wet delay
Surface temperature
Factor Π
Fig. 1: Pixel geometry (ECMWF fields) and rays from 24
satellite positions (3 GNSS satellites, 8 positions per
satellite during a 30 min interval) to 10 ground stations
(indicated by squares). A total of 226 rays is shown, as
only rays, that do not leave the reconstruction field
sideward were used for the retrieval. The most slant rays
are at 7° elevation (note the plot aspect ratio of ~ 1:22).
Slant integrated water
vapor measurement
Tomographic scheme
y = A·x + ε
y = measurement vector (slant integrated water vapor)
A= design matrix (ray-path lengths within pixels)
x = state vector (water vapor densities within pixels)
ε = measurement error vector
Fig. 2: Excess path delay (black), slant hydrostatic delay
(blue) and wet delay (red) for a refractivity field located
in Florida (centered at 27°N, 80°W). Each bundle of 24
ray-numbers corresponds to ray-paths between the
different satellite positions and one ground station.
Summary and conclusions
Tomographic imaging becomes feasible when ground-based measurements are
combined with spaceborne measurements, which requires co-location of ground
receivers and occultation events. We developed a technique for tropospheric water
vapor imaging, where the ground-based line integral measurements are combined with
an occultation profile employing optimal estimation. Instead of occultations also other
profile data (e.g., from radiosondes) could be used.
The retrieval algorithm was tested by computing different scenarios with the aid of
simulated data. We conclude that the presented retrieval algorithm is capable to
reproduce realistic atmospheric features, like secondary water vapor maxima near the
top of the tradewind inversion.
In areas with low absolute humidities, the occultation accuracy is significantly
affected by the accuracy of the required a priori temperature profiles. A procedure
like the mentioned IWV adjustment should be employed in this case in order to
suppress biases. But even in areas with low absolute humidities, like in Finland, useful
two-dimensional information can be obtained with the presented optimal estimation
In areas with high absolute humidities, variations of the water vapor density are
generally less pronounced, the occultation profile is less sensitive to errors in the a
priori temperature profile, and the retrieval results are generally of good quality.
We are confident that the proposed methodology will find fruitful application to
genuine data and thus contribute to the provision of much needed information on the
regional and global water vapor distribution.
T.R. Emardson et al., J. Geophys. Res., 103, 1 807-1 820, 1998.
J. Duan et al., J. Appl. Met., 35, 830-838, 1996.
R. Ware et al., Geophys. Res. Lett., 24, 417-420, 1997.
C. Rocken et al., J. Geophys. Res., 102, 29 849-29 866, 1997.
C.D. Rodgers, J. Geophys. Res., 14, 609-624, 1976.
 2000 by IGAM/UG
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF