EUMETSAT and GRAS

1
EUMETSAT and GRAS
C. Marquardt, A. von Engeln, Y. Andres, F.
Sancho Thanks to many, many more people
GSN/ESOC, GRAS SAF (DMI, MetO, ECMWF, IEEC), UCAR
2
Content

• GRAS instrument
• Some operational aspects
• Raw sampling
• Structural uncertainty
• Infrastructure GRAS SAF, GSN
• Summary

3
GRAS GNSS Receiver for Atmospheric Sounding

• Build by Saab Space (now RUAG, Sweden instrument
  and antennas) and Austrian Aerospace (Austria
  S/W, DSP)
• Mass 30 kg, power 40 W
• 20 MB per orbit / 280 MB per day
• No heritage from the JPL series of RO receivers
  (GPS/MET, CHAMP, GRACE, COSMIC, TerraSAR-X, )
• H/W S/W design AGGA2 until 1999 GRAS until
  2003 (or so)

Launched 19th Oct 2006 switched on 27th Oct
2006 worked out of the box. Bending angles
operational 04/2008, ECMWF 05/08, MetO (N) 07/08
refractivity pre-operational 03/09
4
Distribution of Data

• 650 occultations per day / single instrument
  (sustained - dont use PRN 32 yet)
• Requirement / expectation was 500 but we have
  more GPS satellites than nominal
• Metop is in a sun-synchronous orbit (930 am) by
  design
• always samples the same part of the diurnal cycle

5
Statistics

• 4 weeks (30 Sep 2007 27 Oct 2007) 692 / 627
  occs / day
• Statistics collected from an early version of the
  current GRAS prototype
• Operational statistics similar (but without me
  bugs-)

6
Operational Aspects / Anomalies

• Buffer overflow on rare occasions (5 times since
  launch) a threshold parameter needed to be
  adapted via firmware patch
• Single event of simultaneous tracking loss on the
  zenith channels receiver recovered after a few
  seconds (SAA)
• GPS SV selection for onboard POD based on GPS
  Almanachs instead of Ephemerides
• occasional degradation of onboard navigation
  solution
• single event divergence of the onboard POD
  required navigation solution reset from the
  ground
• GRAS is the instrument with the lowest number of
  instrument anomalies on Metop (by far)

7
Raw Sampling

• Operational GRAS processing currently based on
  geometrical optics
• Raw sampling (or open loop) data not yet
  exploited
• Currently prototyping reconstruction and
  retrieval algorithms for tropospheric data in a
  joint study with ESA / RUAG / DMI GRAS SAF /
  University of GRAZ / GFZ (and DLR for orbits)
• First retrievals to the ground done by DMI
• We expect results to become operational by the
  end of the year
• Demonstration data will be available earlier
• Reconstruction software will be available as OSS

8
Raw Sampling (contd)
C/A
P1, P2
9
Spectra / Radioholography (Rising)
10
Structural Uncertainty
ECMWF bias

• 30 days before Sep 24, 2008
• Operational monitoring at ECMWF
• Similar results at GRAS SAF, UK MetOffice, Meteo
  France
• Altitude region where RO is considered to be most
  accurate (and unbiased)

GRAS (EUM) CHAMP/GRACE (GFZ)
COSMIC (UCAR)
(figure courtesy Sean Healy, ECMWF, from
operational monitoring)
11
Structural Uncertainty (contd)

• Processing COSMIC excess phases (UCAR offline
  processing, downloaded 01/2009) with GRAS
  prototype
• identical parameter settings for both retrievals
  (but POD differencing independent)
• UCAR processed bending angles as reference
• 30 Sep 28 Oct 2007
• 37157 (53813) COSMIC profiles
• 18577 GRAS profiles
• Compared w/ colocated ECMWF co-located 12 hourly
  forecasts
• Geometrical processing only DONT look into the
  troposphere

Robust statistics (Tukeys biweight) for
objective Q/C
12
Structural Uncertainty (contd)

• Systematic biases between 20 and 30 km understood
  by now (implementation of excess phase filter)
• Updates in the UCAR processing will address this
  in the near future
• Biases above 30 km remain to be understood
• Yes, there is structural uncertainty in RO, but
• even significantly different GPS instruments
  agree exceptionally well when being processed
  consistently
• Intercomparing independent processing results
  will allow to quantify (and possibly resolve)
  these issues
• Effects are tiny a few 0.1 (corresponding to a
  few tenth degrees)!

Robust statistics (Tukeys biweight) for
objective Q/C
13
Upper level noise (bending angles)

• Same experiment
• upper level (60 80 km) noise estimates against
  CIRA/MSIS
• Used to demonstrate conformance with
  requirements max. abs. error lt 1 µm,
  one sigma 0.65 µm
• COSMIC and CHAMP data processed with same
  settings as GRAS data (in particular filter
  settings) to get a comparable relative noise
  estimate
• Lower noise due to USO zero differencing

14
Upper level noise (excess phase)

• Generalized Cross Validation for spline fitting
  over first 20 seconds
• RMS from fit as proxy for (thermal) noise and
  differencing errors

15
GRAS Meteorology Satellite Application Facility

• GRAS SAF is a joint effort between
• DMI - operational level2 processing (N, P, T, q,
  climate products), advanced algorithms (CT, FSI)
• MetO - Data Assimilation, ROPP (forward operators
  for RO data assimilation, data thinners, format
  converters)
• ECMWF - Data Assimilation, Re-analyses
• IEEC Reflections, data quality
• http//www.grassaf.org
• for monitoring, data S/W, user services,

16
GRAS SAF (contd)

• Refractivity product declared pre-operational
  in March 2009
• Still some issues w/ refractivity biases above 30
  35 km due to statistical optimisation
• Improved refractivity and additional level 2
  products (p, T, q) expected in June/July, based
  on ROPP
• Climate products (gridded data, monthly means
  etc.) in preparation
• There will also be an offline product of GRAS
  data for use in climate applications (better
  orbits, improved algorithms)

Comparison w/ ECMWF (breakthrough and
threshold)
17
GRAS Ground Support Network (GSN)

• Provides auxiliary data for GRAS processing
• designed, developed and validated by the
  Navigation Support Office at ESOC
• global network of 40 on-ground GPS receiver
  stations delivering data to a processing centre
  located at ESOC through subcontractors (NRCan,
  Fugro, GFZ, ESOC)
• Data collected and processed at ESOC HQ in
  Darmstadt
• GPS orbits, clocks, EOPs, network data for
  differencing (not used at present) in NRT
  sufficient for NWP applications

Primary set of GSN ground station network
18
GRAS GSN (contd)

• Evolution over the next months
• Port to state-of-the-art NAPEOS
• Identical software as used by ESOCs IGS
  processing (but still differences due to
  different input data etc.)

3D-RMS
19
Summary

• GRAS has been performing excellently since
  Metops launch
• No major anomalies
• Only a single firmware update so far (and none
  foreseen) to fix a parameter setting
• So far, data quality continues to exceed
  expectations / requirements
• Starting to look into raw sampling data
• Excellent consistency between closed loop and raw
  sampling data
• current emphasis
• advanced algorithms taylored for GRAS data,
  navigation bit removal
• Joint study w/ RUAG, DMI/GRAS SAF, University of
  Graz, and further collaborations with GFZ Potsdam
  and UCAR/COSMIC ongoing
• first data being analysed

20
Summary (contd)

• Structural uncertainty
• 2009 Re-engineering of the operational GRAS
  processing software to
• extend GRAS data into the troposphere (raw
  sampling)
• improve orbit quality stability
• develop multi-mission capability
• reprocess existing data set with improved
  algorithms
• GSN data quality will improve
• Operational bitgrabber network being considered
• SAF Level 2 products will become available