GPSRO Error Characterization: occultation data recorded in open-loop and closed-loop mode.

Download Report

Transcript GPSRO Error Characterization: occultation data recorded in open-loop and closed-loop mode.

GPSRO Error Characterization:
Analysis of RO measurement errors based on SAC-C radio
occultation data recorded in open-loop and closed-loop mode.
Martin S Lohmann
1
JCSDA Science workshop April 2005
This presentation
2

CDAAC (COSMIC) processing overview

Summary of error estimation techniques
– Stratosphere upper troposphere (GO-region)
– Mid and lower troposphere (RH)

SAC-C open-loop and closed loop error characteristics
– Examples of error profiles
– Error statistics and comparison with ECMWF analysis
– Error distributions
– Error correlation lengths

Summary slide
JCSDA Science workshop April 2005
Radio occultation principle
Ionosphere

GNSS
Neutral
Atmosphere
Earth
LEO
GNSS = Global Navigation Satellite System
LEO = Low Earth Orbiter
 = Bending angle
Signal frequencies: f1 = 1.57542 GHz & f2 = 1.22760 GHz
p
e
Ne


1

k

k

C
1
2
Refractive index of medium:
T
T2
f2
3
JCSDA Science workshop April 2005
Open loop vs. Closed Loop - outline
4

GPS radio occultation signals can be tracked using either traditional so-called
closed loop/phase locked loop (PLL) tracking or so-called open loop tracking
(OL)

PLL is based on a feed-back from the tracked signal itself which may result in
tracking errors. PLL is performing (almost) optimal noise filtering when tracking
correctly
Severe tracking errors are common in GPS-MET, CHAMP and SAC-C RO data
particularly in the lower troposphere. Tracking errors require sophisticated QC
procedures to remove

OL tracking does not involve any feed-back mechanisms. Consequently, OL RO
are not affected by tracking errors - but can be more noisy

COSMIC and SAC-C (in open loop mode) switch to OL tracking below approx.
10 km
JCSDA Science workshop April 2005
CDAAC RO processing - overview
Occultation processed by CDAAC can be divided into two separate
regions where different data processing strategies are applied:
1. GO – region: processing is based on geometrical optics (GO). Bending
angle and refractivity profiles are retrieved from the GPS L1 and L2 signal
phases. Statistical optimization (use of climatology) is applied in the retrieval
of refractivities. Data are smoothed over approx. 1 km
This region covers the height range from the top of the occultation to the
lowest height for which the L2 signal is being tracked
2. RH – region: processing is based on radio holographics (RH) and bending
angle and refractivity profiles are retrieved from the L1 phase and amplitude
using FSI
The RH region extends from the lowest height to which the L2 signal can be
tracked down to the lowest point in an occultation. Data are smoothed over
approx. 100 m
5
JCSDA Science workshop April 2005
RO measurement errors
GO Region:
 Climatology used for retrieval of refractivity
 Background/ionospheric noise
 Tracking errors (closed loop)
RH Region:
 Early signal truncation (closed loop)
 Background/ionospheric noise
 Small scale horizontal variations
6
JCSDA Science workshop April 2005
Error estimation techniques - outline
7

GO region: CDAAC (COSMIC) error estimation is based on (Lohmann 2005):
Below the E-layer the magnitude and structure of the bending angle
measurement errors are fairly uniform [Kursinski et al., 1997; Kuo et al., 2004]
Consequently, the bending angle errors can be estimated from high altitude
differences between a climatology and the observations
Errors in the climatology used for SO are estimated from differences between
measurements and climatology in the lower part of the GO region

RH region: CDAAC (COSMIC) error estimation is based on (Lohmann 2006):
Measurement errors are estimated by mapping fluctuations in the FSI-amplitude
to bending angle errors
In the lower troposphere where the RO signals are very noisy, an alternative
technique is applied where small scale bending angle fluctuations are considered
as errors
JCSDA Science workshop April 2005
SAC-C occultation, April 30, 2005, 14.34 UTC, 24S- 82W.
8
JCSDA Science workshop April 2005
Estimated errors vs. SAC-C - ECMWF differences
Data from March 16 to May 16 2005
9
JCSDA Science workshop April 2005
Error distributions low latitudes
10
JCSDA Science workshop April 2005
Error distributions mid latitudes
11
JCSDA Science workshop April 2005
Error distributions high latitudes
12
JCSDA Science workshop April 2005
Error autocorrelation functions (1)
13
JCSDA Science workshop April 2005
Error autocorrelation functions (2)
14
JCSDA Science workshop April 2005
Summary
Title: JCSDA GPSRO assimilation
Purpose: GPSRO error characterization
Progress so far:
 Implementation of dynamic error estimation (will be included in the next CDAAC
operational update)
 Extensive analysis of GPSRO measurement error characteristics (error profiles,
error distributions, and error correlation lengths)
Future plans:
 Post launch fine-tuning of QC and data-processing
 Better understanding of model minus observation differences
 Investigating the possibility of filtering out gravity waves and other atmospheric
structures which are not included in NWP model fields
15
JCSDA Science workshop April 2005