1. Challenges of Radio Occultation Data Processing
in the Neutral Atmosphere
S. Sokolovskiy
COSMIC Retreat, October, 2011

2. Upper stratosphere and lower troposphere are the regions of
maximum errors and uncertainty of the GPS RO inversions
In the lower troposphere:
the signal reduces below noise level
in terms of the amplitude
Additive noise - main error source
In the upper stratosphere:
the signal reduces below noise level
in terms of the exc. phase (Doppler)
Multiplicative noise - main error source

3. Main problem of RO in LT - water vapor:
- strong mean defocusing
- fluctuation of phase and amplitude
- accurate connection of phase (a fundamental observable of RO) is not possible
high latitudes, winter:
- low water vapor, no problem
- RO signal blocked by surface
before defocusing to noise
- phase stays connected
- low sensitivity to noise
low latitudes:
- strong defocusing & fluctuation
- RO signal gradually descends
below noise level
- connection of phase not possible
- inversion results depend on
cutoff height and noise level

4. Uncertainty of wave optics inversions in LT
cutoff height: km; km
solid - as is; dashed - added noise (17V/V)
tropics
polar
winter

5. Transforming RO signals from time - frequency to
bending angle - impact parameter representations
polar
tropics
noise
cutoff
Non-spherically symmetric
irregularities of N in LT
broaden local spectrum of
WO transformed RO signal
and make BA sensitive to
cutoff height and noise

6. Testing of deep OL tracking (down to HSL=-350km) on FM-1 on Oct
1) Down to what heights the LT RO signals can be observed?
2) What other signals (contaminant to RO) are observed?
Interfering GPS signals were
first found in Metop-GRAS data,
Bonnedal et al., 2010
Interfering signal from GPS
PRN 23
The height from
which the deep
signal arrives
RO signal propagated through
LT, observed down to -300 km.

7. Main sources of RO inversion errors (including biases) in LT
No significant errors in high latitudes, in winter
Largest errors in tropics; reason: water vapor (defocusing, fluctuation)
- receiver tracking errors (resolved with open loop tracking)
- superrefraction:
--- bias in refractivity (bending angle not biased)
--- constrained inversions of BA to refractivity
--- direct assimilation of BA in LT (difficult, large representativeness errors)
- spread asymmetric spectrum of RO signal in LT
- effect of noise
- dependence on the length of RO signal
- horizontal gradients (negative bias?)
- how deep the signals shall be used ?
(can deep signals be caused by spherically symmetric N)
What can be done to reduce inversion errors in LT?
- reflected signals
- accurate measurement of delays

8. Ionospheric correction: LC=Ca*La-Cb*Lb
where Ca=fa^2/(fa^2-fb^2); Cb=fb^2/(fa^2-fb^2)
For La=L1= GHz, Lb=L2=1.2276GHz, Ca=2.546, Cb=1.546
For La=L2=1.2276GHz, Lb=L5= GHz, Ca=12.252, Cb=11.252
Errors of the ionospheric correction:
- ray separation at different frequencies:
- higher order terms of dependence of refractivity on frequency
- diffraction effects by small-scale irregularities
GPS
LEO
irregularities L1 L2

9. eliminated) by combining L1 and L2 BA at the same impact parameter
Large-scale error due to ray separation is substantially reduced (but not
eliminated) by combining L1 and L2 BA at the same impact parameter
Residual error can be estimated by ray-tracing, using a model of electron density,
and applied for correction of ionosphere-free bending angle
An alternative approach (more simple but less accurate): a non-linear regression on
L4=L1-L2 bending angle

10. Bending angle bias correction for solar max. 2002 and solar min. 2007
For improved accuracy, direction of magnetic field must be taken into account
(higher order terms)

11. L2 LEO GPS L1 irregularities
Small-scale ionospheric irregularities introduce fluctuating error times
larger than the large-scale (bias). Main error source at heights > 30 km.
Reasons:
- ray separation
- diffraction effects
GPS
LEO
irregularities L1 L2
Left occ:
larger bias error
Right occ:
larger fluct. error

12. Different types of ionospheric scintillation
A - no scintillation
B,C,D - scintillation in F-layer
E,F,G - effect of E(Es) layer,
can results in large inversion errors at km
Scintillation from E(Es)
layer is observed ~10 times
more often than from F-layer
Effect of Es layer on RO signals

13. Localization of ionospheric irregularities by back propagation
observation trajectory
(phase & amplitude)
solving Helmholtz equation
in a vacuum, given boundary condition Y X
F-layer
E-layer

14. Localization by BP requires elongated irregularities of a given orientation.
Can BP reduce the errors of the ionospheric correction without the
knowledge of orientation of irregularities?
L1 bend. ang.
L2 bend. ang.
LC bend. ang.

15. In Gras-METOP data scintillation from F-layer is observed more
frequently than from E(Es)-layer
(due to uneven sampling in LT)
An example of Gras-
METOP occultation
with localizable
scintillation
in F-layer
Ionospheric correction
see next slide

16. Ionospheric correction after BP propagation of RO signal to location
of irregularities (+2200 km) allows to significantly reduce the residual
Errors of the ionospheric correction
Unfortunately, this improvement is not possible for all occultations

17. Can reduction of the difference in the two frequencies reduce
the diffraction-related errors of ionospheric correction? NO
L1= GHz, L2=1.2276GHz, Ca=2.546, Cb=1.546
L2=1.2276GHz, L5= GHz, Ca=12.252, Cb=11.252

18. Reduction of noise propagation in inversion of bending angle
Current approaches:
Optimization of BA
1st guess does not
depend on obs. BA
1st guess depends
on obs. BA
1st guess:
climatology
1st guess:
NWP model
1st guess:
fitted exponent
1st guess:
fitted climatology
GFZ
WegC
JPL
DMI, UCAR
Abel Inversion
Generalization: optimal estimation of the state vector (P,T) by including
ancillary information or constraints in the stratosphere
- parametrization of T between mesopause and stratopause
(incl. stratopause height as a parameter);
- including T from NWP or other satellite observations;
- including (T,P) from NWP