Environment Canada

1
Environment Canada

• CMC Monitoring of GLFE TAMDAR Data

Gilles Verner, Yulia Zaitseva, Réal Sarrazin /
Gilles Fournier Canadian Meteorological Centre /
AMWSD
2
Outline

• Background on CMC models and monitoring
• Status of Canadian AMDAR Program Development
• Plans for the Future
• Monitoring of AMDAR data at CMC
• Monitoring of GLFE-TAMDAR at CMC
• Conclusion and Discussion

3

• Background on CMC models and monitoring

4
CMC Operational Models
GEM model used for both global and regional
versions 4D-var assimilation for global system as
of March 15, 2005 Regional system still using
3D-Var
Global Model
Regional Model

• Variable resolution grid
• Resolution of .1375º (15 km)
• 58 eta levels
• Kain-Fritsch scheme
• Sundqvist stratiform scheme
• ISBA surface module with
• soil moisture pseudo-analysis
• (error feedback, no data)
• 48-hour forecasts (00Z -12Z)
• Cut-off of T1h40

• Uniform grid
• Resolution of .9º (100 km)
• 28 eta levels
• Kuo convection scheme
• Sundqvist stratiform scheme
• Force-restore surface module
• with climatogical soil moisture
• 10 day forecasts at 00Z
• and 6 day forecasts at 12Z.
• Cut-off of T3h00

5
Distribution maps of aircraft observations
assimilated in 6-hour period 3D-VAR analysis
CMC
6
Distribution maps of aircraft observations
assimilated in 6-hour period 4D-VAR analysis
The amount of aircraft data assimilated in 4D-VAR
has tripled compared to 3D-VAR!
CMC
7
Data assimilated in 3D-VAR Analysis
Global - Monthly mean number of assimilated
observations per 24 hours over 6h- assimilation
window 3D-VAR analysis (green color aircraft
observations).
CMC
8
Data assimilated in 4D-VAR Analysis
Global - Monthly mean number of assimilated
observations per 24 hours over 6h- assimilation
window 4D-VAR analysis (green color aircraft
observations).
CMC
9

• Status of Canadian AMDAR Program Development

10
Development with AC Jazz

• Largest Regional with 67 DHC-8s and
  increasing fleet of CRJs to 73 by March 2006
• AMDAR started with 21 AC Jazz DHC-8 100 in June
  2002 but T bias issue
• 24 AC Jazz CRJs reporting good T and wind data
  on GTS (73 by March 06)
• 24 AC Jazz upgraded DHC-8 reporting good T and
  wind data (67 Nov 05)

• Canadian AMDAR data assimilated at CMC
• distributed on GTS since 4 Jan 05
• displayed on FSL web since 12 Jan 2005
• CC operational since 3 Feb 2005

11
Sabre CC System
12
FSL web-Canadian data, 24 hrs, 5 Apr 05
13
Aircraft Profiles in Real-Time
14
Development with First Air

• First Proof Of Concept system about to be tested
  on a test B727
• If test is positive and funds available, proceed
  with implementation on 8 aircraft by 31 March 06
  and 6 in FY06/07

15
Historical Background

• 4-phase development contract began in mid-2002
• Phase 1 feasibility analysis (completed by 31
  March 2003)
• not economically and technically possible to
  upgrade each aircraft
• TAMDAR selected due to its promise to be easily
  adaptable to various aircraft configurations and
  requiring minimal certification
• Phase 2 development of POC ISAT/TAMDAR/Internet
  (completed in Fall 2003)
• Phase 3 POC system testing on a B727 (most of
  the delays - hope to be completed by 30 June
  2005)
• certification by FAA and then Transport Canada
  generated significant delays
• lots of unexpected technical problems (GPS, data
  rates, calibration, First Air/Skytrack/AirDat
  priorities)
• capital procurement funding returned twice due to
  these delays
• Phase 4 deployment on 15 aircraft (8 in
  FY05/06 6 in FY06/07)
• if test is positive and funding available
• test to be difficult as, contrarily to GLFE, not
  much data in the north
• calibration request heavier than anticipated
• wind quality in the north is a big unknown
• a lot of the QC moved to the ground processing
  centre

16
Projected Weekly Ascents/Descents
Notes 1. Includes expected CRJ and DHC-8
operated by Jazz and 15 aircraft operated by
First Air 2. Does not cover WestJet and Air
Canada 3. Canadian North would add 30 more data
in North 4. AFIRS/UpTime would be deployed to
fill holes
17
Development with AFIRS/UpTime

• AFIRS Automated Flight Information and
  Reporting System
• Independent datalink system for small airlines
  that cannot afford ACARS
• Per flight hour data fees No upfront costs to
  clients
• Partnership with TCs Flight Data Monitoring
  (FDM) program
• AMDAR capability was developed and tested on 3
  HawkAir DHC-8s operating in BC (T-bias issue)

• AMDAR system based on AMS AFIRS expected to be
  on all 5 B737 aircraft from Canadian North by 30
  June 2005
• A dedicated T/RH sensor integrated to AFIRS is
  being investigated

18

• Plans for the Future

19
Plans for the Future

• Impact studies of Canadian AMDAR data by CMC and
  Canadian operational forecasters(?) in FY05/06
• On-going activities
• Internal development (CMC)
• AC Jazz comms
• First Air comms TAMDAR LCM
• AMS AFIRS/UpTime comms (Canadian North, HawkAir,
  etc.)
• WestJet, Air Canada comms
• LCM for required non aircraft critical systems
• Remaining development activities
• Implement on 15 First Air aircraft
• Expand coverage through AFIRS/UpTime
• AMDAR development on WestJet (B737 aircraft)
• AMDAR development on Air Canada Embraers ERJs
• Business Case to EC for the measurement of
  humidity
• Aviation-related (icing, turbulence) BC to NAV
  CANADA

20
BC to NAV CANADA

• Objectives contribution of NAV CANADA sought on
  development and operation of
• AMDAR turbulence reporting capacity
• AMDAR icing reporting capacity
• AMDAR Program on-going communication costs
  associated with expanding AMDAR coverage

21
AMDAR VS GEOSS

• AMDAR meets all global GEOSS requirements
• Affordable
• Expandable
• Sustainable
• Global coverage
• International standards
• Can target observations
• Best global in-situ tropospheric data for
  satellite calibration
• Air Quality Sensing Load?
• Aircraft mesoscale network filling hi-res plume
  dispersion model in case of a NCB attack?

22

• Monitoring of AMDAR data at CMC

23
Aircraft Sensor Monitoring

• Meteorological Centres such as CMC that run
  Numerical Weather Prediction (NWP) models can
  monitor the performance of aircraft sensors used
  in AMDAR on a continuous and real-time basis
• Monitoring based on observed minus first guess
  values (innovations), as well as data rejection
  statistics, extracted from operational data
  assimilation system
• Monitoring is performed for individual aircrafts
  as well as by AMDAR programs (e.g. E-AMDAR, GLFE,
  etc).
• Time evolution of innovations, as well as their
  statistical distribution are extremely useful
  tools

24
QC techniques at CMC

• In CMC 3D and 4D-Var, data QC based on 2 checks
• A simple background check (comparison with first
  guess, data are rejected if departure from first
  guess is larger than pre-specified limits (4-5
  times the normalised std deviations). This is
  used to identify large (or gross) errors
• A more sophisticated variational quality control
  which is applied during the minimisation process,
  taking into account the consistency of the
  observations with other observations as well as
  the first guess and the final analysis. QC
  decisions can (and do) change during the
  minimisation process

25
Ex. T Bias on DHC-8 Aircraft
Aug 02
Unacceptable mean T-bias over 2C from DHC-8 using
original OEM temperature probe
CMC
26
Ex. T Bias on DHC-8 Aircraft
Aug 02
Significant change in TT/UV biases probes
changes by Jazz in Dec 04!
CMC
27
Ex. T Bias on DHC-8 Aircraft
Density plot of innovations of temperature, all
data for month of October 2004. Note the known
temperature bias of the DHC-8
CMC
28
Ex. T Bias on DHC-8 Aircraft
Density plot of innovations of temperature, all
data for month of January 2005. Note that the
temperature bias of the DHC-8 is gone!
CMC
29
Ex. DHC-8 Aircraft Wind Monitoring
Scatter plot for wind, all data for month of
March 2005. Note a few bad values when
forecasting light winds!
CMC
30
AC Jazz Data assimilated at CMC
Impact of 4D-Var
CMC
31

• Monitoring of GLFE-TAMDAR at CMC

32
Monitoring of GLFE TAMDAR at CMC

• Data in BUFR format obtained from AIRDAT ftp
  server and processed like other AMDAR
• Special care was taken to properly interpret
  quality flags which are present in the BUFR
  files TAMDAR data flagged as SUSPECT or BAD were
  NOT included in the monitoring, but are available
  in the database. Counts on how many data are
  flagged.
• Monitoring done for all data as well as for
  individual aircraft.
• Tables of suspect data generated on a monthly
  basis using the standard WMO criteria
• Results available on a monitoring web site
  (intranet)
• Test restricted to temperature and wind

33
GLFE Data Received at CMC
GLFE observations decoded by CMC. About 4700
observations from all levels, over a 6-hour
window centered at 18 UTC on 04 April 2005.
CMC
34
GLFE Data Received at CMC
Time series of the amount of data received, 25
day period.
CMC
35
Monitoring of GLFE TAMDAR Data at CMC
Innovations of MVD and speed bias, all data with
good flag only. Note speed bias.
CMC
36
Monitoring of GLFE TAMDAR Data at CMC
Innovation of temperature, all data with good
flag only. Note occasional larger deviations, but
biases remain small.
CMC
37
Monitoring of GLFE TAMDAR Data at CMC
Innovation of temperature, all data with good
flag only, month of February. Note more frequent
larger deviations, but biases remain small.
CMC
38
Monitoring of GLFE TAMDAR Data at CMC
Bad data
Density plot of innovations of temperature, all
data with good flag only, month of February. Note
some very bad data (large deviations) but with
good flag. These bad data are flagged by CMC
background check.
CMC
39
Monitoring of GLFE TAMDAR Data at CMC
Scatter plot for wind, all data for month of
March 2005. Note some bad wind data (larger
deviations) but with good flag. These bad data
are affecting the overall statistics.
Bad data
CMC
40
Single Aircraft GLFE-0217
Density plot of innovations of temperature for
aircraft GLFE-0217, all data with good flag only,
month of March. Note some very bad data (large
deviations) but with good flag
CMC
41
Single Aircraft GLFE-0205
Density plot of innovations of temperature for
aircraft GLFE-0205, all data with good flag only,
month of March. TAMDAR TT data is generally of
very good quality.
CMC
42
Single Aircraft GLFE-0205
Scatter plot for wind for aircraft GLFE-0205, all
data with good flag only, month of March. TAMDAR
wind data is generally of very good quality. Some
positive speed bias a small concern.
CMC
43
Monitoring criteria for Suspect Aircraft

• March 2005 TAMDAR
• Pressure Categories (hPa)
• LOW PRESS 701 - SFC
• MID PRESS 301 - 700
• HIGH PRESS 300 - 100
• ID is the aircraft tail number
• NA is the total number of available
  observations
• NE is the total number of erroneous
  observations
• NR is the number of rejected observations
  
• NG is the number of gross observations
  excluding erroneous data
• NC is the number of exactly calm winds
  excluding erroneous data
• TBIAS is the temperature bias for non-gross
  temperatures and non-erroneous data
• TRMS is the RMS temperature difference
  excluding gross errors and erroneous data
• SBIAS is the speed bias for non-gross winds
  and non-erroneous data
• WRMS is the RMS wind difference excluding
  gross errors and erroneous data
• Selection criteria num obs LOW20,
  MID50, HIGH50

44
Results as a table

• BUFR FORMAT TEMPERATURE OBSERVATIONS
• SUSPECT TEMPERATURES
• ID ELEM LEVEL NA NE NG NR
  TRMS TBIAS
• GLFE0238 TEMP 301-700 156 6 32
  36 1.9 -1.1
• GLFE0283 TEMP 301-700 338 106 61
  120 1.1 -0.8
• GLFE0217 TEMP 301-700 2392 163 140 196
  1.5 -0.6
• GLFE0251 TEMP 301-700 2034 110 82
  142 1.2 -0.4
• GLFE0238 TEMP 701-SFC 213 92 56
  109 1.9 -0.2
• GLFE0283 TEMP 701-SFC 509 119 113 201
  4.1 0.6
• GLFE0217 TEMP 701-SFC 3757 195 297 585
  3.7 0.6
• GLFE0242 TEMP 301-700 616 9 0
  13 3.3 3.0
• NON-SUSPECT TEMPERATURES
• ID ELEM LEVEL NA NE NG
  NR TRMS TBIAS
• GLFE0247 TEMP 701-SFC 647 456 0
  342 1.9 -1.2
• GLFE0262 TEMP 301-700 2131 63 0
  16 1.4 -1.0

45

• Conclusion and Discussion

46
Conclusion and Discussion

• Summary of monitoring results
• TAMDAR data generally of good quality. Some
  concern about a small positive wind bias
• Some obviously bad data are making it to the BUFR
  files and are corrupting the overall statistics.
  This is affecting a few aircraft (for February,
  GLFE 217,225, 240,244,248,249,253,255,270,271,275,
  279 and 287.
• These bad data are transmitted with a good
  quality flag (data with the bad flags are not
  used).
• A more stringent QC at the source should be
  considered to remove these bad data
• These bad data are usually identified by the NWP
  QC processes. This would prevent their
  assimilation.
• Monitoring by NWP process important and useful to
  identify issues with data