1 AMDAR Quality Assurance Bradley Ballish NOAA/NWS/NCEP/NCO/PMB SSMC2/Silver Spring 23 March, 2009.
-
Upload
sabina-white -
Category
Documents
-
view
221 -
download
6
Transcript of 1 AMDAR Quality Assurance Bradley Ballish NOAA/NWS/NCEP/NCO/PMB SSMC2/Silver Spring 23 March, 2009.
1
AMDAR Quality Assurance
Bradley BallishNOAA/NWS/NCEP/NCO/PMB
SSMC2/Silver Spring23 March, 2009
2
Outline• Monthly reports
• Examples of data quality control (QC) problems
• Comparison of some aircraft temperatures, wind and moisture data in North American area
• Proposed aircraft temperature bias corrections and related issues
• Summary
3
Regular Monthly AMDAR Reports
• Based on a WMO meeting at the ECMWF in June 2002, NCEP prepares monthly aircraft monitoring reports at website:
• http://www.ncep.noaa.gov/pmb/qap/amdar/• These standard monthly reports are not frequent
enough in time, do not have track-checking or stuck data summaries and do not have accent and descent statistics in most parts
• The WMO Integrated Global Observing System (WIGOS) Pilot Project for AMDAR suggests regional centers QC AMDAR data before transmission on the GTS
• This will require much more frequent updates than monthly reports
4
Japanese Data in Monthly Reports
• In the NCEP AMDAR report for February 2009, the Japanese data looked good
• Of 274 Japanese aircraft reporting data, only 7 had suspect temperatures:
• Units JP9Z4U44, JP9Z4Y4X, JP9Z4Y79, JP9Z4YVV, JP9Z5859, JP9Z585Z and JP9Z5Y79 had warm biases
• No units had suspect winds!• There were about 100 minor track-check
errors, see example on next page
5
Track-Check Error Example
Aircraft Data for Unit JP9Z58XZ
For 00Z 16 March 2009
Time-Days Lat Lon Press
16.07153 27.40 125.00 196.8
16.07639 34.82 140.37 461.7
16.07708 29.53 127.97 196.8
16.11944 34.75 140.28 435.2
16.12083 32.05 132.23 196.8
16.14306 35.27 140.70 558.1
16.14444 35.48 140.78 609.7
16.22917 35.40 139.90 290.1
16.26806 34.32 133.58 300.9
16.28750 33.52 130.43 300.9
Locations and pressures arechanging too fast with timebut all data are close to modelbackground
All raw data received at NCEP haveonly header KAWN – US Air Forcenot RJTD as expected Additional examples can be provided
6
Aircraft Monitoring Example
• On 9 August 2006, aircraft EU3102 started to show a large temperature bias from 300 hPa up compared to the background
• The spurious bias was so large that few spurious temperatures passed QC
• The bias was so large that the aircraft was probably wasting fuel
• If the airlines could check a website with this information, such problems could be found and fixed much sooner
7
Temperature Bias for unit EU3102 300 hPa up August 2006
-2
0
2
4
6
8
10
1201
00
0106
0112
0900
1100
1118
1206
1212
1300
1306
1318
1400
1506
1512
1518
1612
1618
1706
1712
1718
1800
1806
Run Day and Time
Tem
per
atur
e B
ias
8
Aircraft Track-check Example• On 11 August 2006, aircraft AFZA01 was flying from the
southeast to northwest with roughly several minutes between reports
• Three groups of reports are shown, with groups 1 and 3 with correct locations and group 2 with all reports about 12 degrees too far north
• The blue numbers are vector wind differences to the guess, with group 3 having large differences that all passed QC
• Flying from the end of group 1 to the start of group 2 is an impossible distance in several minutes
• This is a tough example for current QC codes to correctly process as group2 can track-check with itself
• This problem with South African aircraft has lasted over a year
• Examples of solo track-check errors are common
9
1
3
2Blue numbers are vector winddifferences of observed windsminus model background
10
Aircraft Temperature Observation Count
Comparison for NA area
• An impact test adding TAMDAR and Canadian AMDAR data at NCEP did not have positive impact, so here we examine this data
• The next slide compares the average number of different types of temperature counts to the nearest mandatory pressure level per GDAS model run in June 2008 for North America (NA)
• Counts for Radiosondes, ACARS, TAMDAR and two types of Canadian AMDAR are compared
• Wind observation counts (not shown) were found to be nearly identical to temperature counts
• Clearly the aircraft counts out number those from sondes
• The two main types of Canadian aircraft are labeled CRJ and DHC-8
11
Average Temperature Counts per Run NA Area June 2008
0 1000 2000 3000 4000 5000
1000
925
850
700
500
400
300
250
200
Pre
ssur
e in
hP
a
Counts
CRJ
DHC-8
TAMDAR
ACARS
Sondes
Sondes have low counts relativeto large ACARS counts
12
Temperature Bias Comparison
• The next slide compares the average temperature bias of different types of observations to the nearest mandatory pressure level per GDAS model run in June 2008
• Biases for sonds, ACARS, TAMDAR and two types of Canadian AMDAR are shown
• Clearly the aircraft temperatures are generally warmer than those from sonds (as found for ACARS and AMDAR, Ballish and Kumar (BAMS, Nov 2008))
• The DHC-8 aircraft have the warmest bias
13
Temperature Biases Versus Guess NA Area June 2008
-1 -0.5 0 0.5 1 1.5
1000
925
850
700
500
400
300
250
200
Pre
ssur
e in
hP
a
Temperature Bias in Degrees
CRJ
DHC-8
TAMDAR
ACARS
Sondes
Sonds are coldcompared toaircraft
14
Temperature Bias vs POF for Canadian AMDAR Data
• In the following slide, the temperature biases for Canadian AMDAR type DHC-8 are shown vs the phase of flight (POF)
• This aircraft type has generally warm biases that vary with the POF
• Here the biases vary considerably with the POF
15
Temperature Biases Versus Guess DHC-8 Aircraft NA Area June 2008
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
1000
925
850
700
500
400
Pre
ssur
e in
hP
a
Temperature Bias in Degrees C
TOTAL
LEVEL
DESCENT
ASCENT
Ascent vsdescent is large
16
Speed Bias vs POF for Canadian AMDAR Data
• In the following slide, the wind speed biases for Canadian AMDAR type CRJ are shown vs the POF
• This aircraft type has speed biases that vary considerably with the POF
• In the second following slide, the same is shown for Canadian aircraft type DHC-8
• Here the speed biases vary even more with the POF
• At the WIGOS February 2009 meeting, it was noted that the CANADIAN AMDAR data are less accurate in high latitudes due to using magnetic, rather than GPS navigation
17
Speed Biases Versus Guess CRJ Aircraft NA Area June 2008
-1.5 -1 -0.5 0 0.5 1 1.5 2
1000
925
850
700
500
400
300
250
Pre
ssur
e in
hP
a
Speed Bias in m/sec
TOTAL
LEVEL
DESCENT
ASCENT
Ascent vs descent is large
18
Wind Speed Biases for Aircraft Type DHC-8 June 2008 NA Area
-1.5 -1 -0.5 0 0.5 1 1.5 2
1000
925
850
700
500
400
Pre
ssur
e in
hP
a
Speed Bias in m/sec
TOTAL
LEVEL
DESCENT
ASCENT
Ascent vs descent is very large
19
Relative Humidity Bias Comparison
• The next two slides show counts of moisture observations and relative humidity biases differences versus the guess for the North American area in June 2008 for sonds, ACARS and TAMDAR data
• The TAMDAR data (at this time) are mainly in the mid west, yet have higher counts and very good stats versus the guess
20
Average Moisture Observation Counts per Run NA Area June 2008
0 200 400 600 800 1000
1000
925
850
700
500
400
300
Pre
ssur
e in
hP
a
Counts
TAMDAR
ACARS
Sondes
TAMDAR has large counts, butare just in mid-west only
21
Relative Humidity Biases Versus Guess NA Area June 2008
-10 -8 -6 -4 -2 0 2 4 6 8 10
1000
925
850
700
500
400
300
Pre
ssur
e in
hP
a
RH Bias in %
TAMDAR
ACARS
SondesTAMDAR biases maybe better than sonds
22
Proposed Aircraft Temperature Bias Corrections
• Ballish and Kumar BAMS(Nov 2008) studied aircraft temperature biases and proposed bias corrections shown in the next slide for January 2007 for the 15 aircraft types with the largest counts
• In the following slide, the same is shown for non US AMDAR types
• This study did not include TAMDAR or Canadian AMDAR types
23
Proposed ACARS Temperature Bias Corrections January 2007
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.675
7-22
2
757-
223
757-
24A
PF
737-
3H4
757-
232
767-
34A
F
A30
0F4-
60
737-
522
MD
-11F
737-
832
MD
-88
757-
251
A31
0-20
3
767-
332
A31
0-32
4
Aircraft Types
Bia
s C
orre
ctio
ns in
Deg
ree
s C
SFC-700 700-500 500-300 300-150
Most correctionsare negative
24
Proposed NUS-AMDAR Temperature Bias Corrections January 2007
-1.5
-1
-0.5
0
0.5
1JP
A32
0-20
0
737-
300
-737 AU
A31
9-10
0 B-
AF
-747
A32
1-10
0
NZ
A34
0-30
0
MD
-11F
A34
0-
Aircraft Types
Bia
s C
orre
ctio
ns
in D
egre
es C
SFC-700 700-500 500-300 300-150
25
Aircraft vs Sond GSI Draws to Temps between 200-300 hPa
# Aircraft >> # Sondes, thus warm aircraft data overwhelms the GSI/GFS system
Aircraft Tdiff (obs-ges)
Aircraft Tdiff (obs-anl)SOND Tdiff (obs-anl)
SOND Tdiff (obs-ges)
26
AMDAR Versus Sond Counts 300-200 hPa
Aircraft
Sonds
Aircraft
Sonds
27
Suru Saha’s website displays model fits to RAOBS in North America showing the GFS analysis and guess maintain a warm bias throughout most of the troposphere that may be related to large numbers of aircraft with warm biases
28
Model Climate Impact from Aircraft Warm Temperatures
• The next slide courtesy of Dick Dee of ECMWF shows the increase in the number of aircraft reports versus time in the ECMWF reanalysis
• The temperature bias of the ECMWF analysis and background seem to be affected by the large increase in the number of aircraft temperatures along with other factors
• The NCEP GSI may have more bias impact as it does not thin aircraft data and its satellite radiance bias corrections are anchored to the analysis as truth as opposed to radiosondes as truth
29
Model Climate Bias Impact From Warm Aircraft Temperatures
Global-mean departures of analysis (blue) and background (red) from radiosonde temperatures (K) at 200hPa, and number of obs/day (x10-4, green)
Global-mean departures of analysis (blue) and background (red) from aircraft temperatures (K) at 200hPa, and number of obs/day (x10-4, green)
30
Summary • The standard monthly AMDAR reports are useful but do not
contain enough information on aircraft data quality• In part due to the WIGOS project, more frequent and complete
quality information will be needed• Improvements are needed in the aircraft track-checking• The TAMDAR data appear to be of useful quality, especially the
moisture• The Canadian AMDAR data show considerable bias differences
with the aircraft phase of flight and will need more effort to assimilate them well
• There is evidence that large numbers of relatively warm aircraft temperatures are impacting model analysis bias
• Improvements are needed in the bias correcting and or use of aircraft temperatures, winds and moisture
• NCEP and the ECMWF are both planning to perform aircraft temperature bias corrections
• It is likely that 4DVAR assimilation is needed to get maximum impact of aircraft data due to their reporting at off times