Many other uses detailed at http:amdar.noaa.gov ... http:

1
Presentation by Earth System Research Lab /
Global Systems Division- Bill Moninger23 March
2009

• Impact of the AMDAR observations to aviation
  weather forecast, public weather service, and
  numerical weather prediction
• request of Mr. Hasegawa from JMA
• Demonstration of ESRL/GSDs real-time display of
  AMDAR dataused by weather services worldwide

2
Bill Moninger, what I look like and where I work
3
What is ESRL/GSD?

• ESRL/GSD is located in Boulder, Colorado
• ESRL has about 500 employees
• GSD has about 200 employees
• We are in the Research branch of NOAA
• (NWS is an Operational branch of NOAA)
• We develop NWP models from global to local scales
• we focus on data assimilation
• we focus on transferring our work to operations
  (NWS)
• We provide data to researchers and operational
  weather forecasters world-wide

4
What we have

• ESRL/GSD operates several large supercomputers
• We gather large amounts of weather data
• including experimental data such as
• WVSS-II
• TAMDAR
• We are a research development organization
• with the flexibility to test new models
• and new data sources

5
Models we run

• Global models (will not be discussed further
  today)
• Mesoscale models
• The Rapid Refresh (RR)
• The High Resolution Rapid Refresh (HRRR)
• The Rapid Update Cycle (RUC)

6

• RR
• 13-km grid
• covers North America
• runs hourly
• HRRR
• 3-km grid
• covers NE US
• soon to cover 2/3 of US
• runs every 15-60 minutes
• RUC
• 13-km grid
• covers US
• runs hourly
• operational for 15 years (in various forms)

7
RUC/RR - backbone for high-frequency aviation
products National Convective Weather Forecast
(NCWF), Icing Potential (FIP), Graphical
Turbulence Guidance (GTG), and the aviation
weather products
Rapid Refresh domain 2009
13km resolution
RCPF
1500 Z 6-h forecast RCPF
Current RUC-13 CONUS domain
Turbulence - GTG
AWC
2100 Z verification
Icing FIP
8
Purpose for the RUC/ Rapid Refresh

• Provide high-frequency mesoscale analyses,
  short-range model forecasts
• Assimilate all available observations
• Focus on aviation and surface weather
• Thunderstorms, severe weather
• Icing, ceiling and visibility, turbulence
• Detailed surface temperature, dewpoint, winds
• Upper-level winds
• Users
• aviation/transportation
• severe weather forecasting
• general public forecasting
• Support from Federal Aviation Administration

Situational Awareness Model
9
Operational Rapid Update Cycle
Hourly updated short-range model run at NCEP
(aviation, severe weather, general forecast
applications)

• Hybrid isentropic coordinate
• Hourly 3DVAR update cycle
• Extensive use of observations
• 13-km horizontal resolution
• Explicit 5-class microphysics

1-hr fcst
1-hr fcst
1-hr fcst
Back- ground Fields
Analysis Fields
3DVAR
3DVAR
Obs
Obs
Time (UTC)
11 12 13
10
Observations assimilated
Cycle hydrometeor, soil temp/moisture/snow plus
atmosphere state variables
Hourly obs in 2008 RUC Data Type
Number Rawinsonde (12h) 80 NOAA
profilers 30 VAD winds
110-130 PBL profiler/RASS
25 Aircraft (V,temp) 1400-7000 TAMDAR
(V,T,RH) 0 - 800 Surface/METAR
1800-2000 Buoy/ship 100- 200 GOES
cloud winds 1000-2500 GOES cloud-top pres
10 km res GPS precip water
300 Mesonet (temp, Td) 7000 Mesonet
(wind) 4500 METAR-cloud-vis-wx
1600 Radar reflectivity 1km
RUC Hourly Assimilation Cycle
11
Commercial aircraft observations - winds and
temperature - recently water vapor,
turbulence
12
Impact of AMDAR data on RUC Forecasts

• Study 1 weekend/weekday skill differences
• Study 2 AMDAR cutoff after 11 Sept 2001
  terrorist attacks
• Study 3 Recent relative impact studies of AMDAR
  and other data sources

13
Study 1 Weekend-Weekday RUC skill differences

• 20,000 fewer reports every 12 hours on weekends
  because package carriers (FedEx and UPS) do not
  fly
• 0000-1200 UTC AMDAR volume average (2001)
• Weekday (Tu-Sa) 35,000 reports
• Weekend (Su-Mo) 15,000 reports
• Result a 7 increase in 3h wind forecast error
  at 200 hPa on weekends
• Study period January-October2001 Stan Benjamin,
  ESRL/GSD

14
3 hr RUC Wind Forecast Errors (with respect to
RAOBs) Weekend (Reduced AMDAR) minus weekday
Jan-Oct 2001
0.35 m/s / 5.0 m/s 7 better forecasts
during weekdays due to more AMDAR reports
15
Study 2 Effect of 11-13 Sept 2001 on RUC Skill

• No AMDAR data due to terrorist attack
• 20 loss of 3h RUC wind forecast skill at 250mb

16
Hourly AMDAR volume2-15 Sept 01(starting 00z 2
Sept)
2-8 Sept 01
Su Mo Tu We Th Fr Sa
9-16 Sept 01
Su Mo Tu We Th Fr Sa
17
Improvement in 3h over 12h wind forecast-
September 2001

• RUC 250 mb
• Wind forecasts
• Verification
• against RAOB data

without AMDAR data, 3-h forecast are no better
than 12-h
11-13 Sep
18
Relative Impact Studies

• These require substantial computer time
• GSD has a research supercomputer on which we run
• multiple retrospective runs, each with a
  controlled change against a standard
• to make detailed tests
• Including TAMDAR evaluation, funded by the FAA

19
Retrospective 10-day experiments

• We used the 2007 version of operational RUC
  model/assimilation software run at 20km
  resolution, with all observations assimilated in
  operational RUC except radar reflectivity
• Two periods August 2007 and Nov-Dec 2006
• Each 10 days long (takes 6 days to run)
• 30 experiments performed on the 06 period

20
Retrospective 10-day experiments (2)

• 13 experiments were completed for the 07 period
• The following data types were excluded
• AMDAR
• TAMDAR
• TAMDAR winds
• TAMDAR rejected aircraft
• Profilers
• NEXRAD VAD wind profiles
• GPS Integrated Precipitable Water (IPW)
• Surface observations (METAR and Mesonet)

21
Temperature relative impact (1)
This shows the impact of each data source shown
for the US Great Lakes Region, during winter
2006, for Temperature forecasts below 6000 ft
(800 mb). AMDAR (red) has the greatest impact of
all data sources investigated for 3h and 6h
forecasts in this region. Surface observations
have the second greatest impact at 3h and
6h. AMDAR has relatively little impact for 12h
forecasts.
Observation types Red AMDAR, including
TAMDAR Blue Profiler Pink NEXRAD VAD Brown
RAOB Blue surface (inc. Mesonets) Green GPS-IPW
Graphs show the error increase when each
observation type is removed.
22
Temperature relative impact (2)
This shows relative AMDAR and TAMDAR impact for
3h Temperature forecasts valid at 0 UTC during
winter 2006. TAMDAR is responsible for about 40
of the total AMDAR impact below 6000 ft. in this
region and during this period. As a specific
example, TAMDAR alone reduces 3-h temperature
errors by 0.5 K at 900 mb (3000 ft.), whereas all
AMDAR data (including TAMDAR) reduces temperature
errors by 1.1 K at 900 mb. More precisely
removing TAMDAR alone increases temperature
errors by 0.5 K, and removing all AMDAR data
increases errors by 1.1 K.
23
Temperature relative impact (3)
This shows the impact of each data source shown
for the Great Lakes Region, during Summer 2007,
for Temperature forecasts. AMDAR (red) has the
greatest impact of all data sources investigated
for 3h, 6h and 12h forecasts in this region.
Surface observations have the second greatest
impact.
Observation types Red AMDAR, including
TAMDAR Blue Profiler Pink NEXRAD VAD Brown
RAOB Blue surface (inc. Mesonets) Green GPS-IPW
24
RH relative impact
Observation types Red AMDAR, including
TAMDAR Blue Profiler Pink NEXRAD VAD Brown
RAOB Blue surface (inc. Mesonets) Green GPS-IPW
Relative Humidity forecast impact for winter
(left) and summer (right), below 6000 ft (800
mb). AMDAR has the greatest impact of all data
sources studied for 3h and 6h in the winter
(left), and for 3h, 6h, and 12h in the summer
(right). TAMDAR is the only AMDAR data source
that provides RH information to the RUC
currently. (We do not yet ingest WVSS-II data.)
25
RH relative impact
This shows relative AMDAR and TAMDAR impact for
3h Relative Humidity forecasts valid at 0 UTC
during winter 2006. In this altitude range (the
lowest 6000 ft.), TAMDAR is responsible for about
60 of the total AMDAR impact for RH in this
region and during this period.
26
Wind impact 3-h wind forecasts (22 - 28 April
2005)
Wind errors are reduced by 1.4 m/s at 200 mb due
to the inclusion of AMDAR data
27
Direct forecaster use of AMDAR data (1)

• Forecasters have direct access to AMDAR data
  through
• ESRL/GSDs web display (to be shown to you soon)
• And through NWS workstations
• (This was covered by Carl Weiss earlier)

• As a radiosonde substitute when there is none
  nearby (Vancouver, CAN and Houston, US)
• To accurately forecast the onset of severe storms
  (near airports with timely flights)
• To forecast and monitor low-level wind shear
• To monitor jet stream location
• To forecast downslope windstorms
• To verify/correct model guidance (Montana, US)
• Fire weather support
• To forecast urban air quality
• Many other uses detailed at
  http//amdar.noaa.gov

28
Direct forecaster use of AMDAR data (2)

• Mountain weather forecasts in support of rescue
  operations (Seattle, US)
• Improved control of aircraft spacing on descent
  (Ft. Worth, US)
• Improved forecast of jet-stream-induced
  turbulence
• Used in aircraft accident investigations (U.S.
  National Transportation Safety Board)
• To initialize a city-scale model used in on-shore
  breeze forecasting (Chicago, US)

29
Ongoing AMDAR observation monitoring

• We generate daily and weekly aircraft-model
  differences
• These are used by us (and others) to monitor
  aircraft data quality
• We automatically generate daily aircraft reject
  lists that are used in our backup and development
  RUC models

30
Typical output from one of our evaluation web
pages
31
Typical output from another of our evaluation web
pages
This shows aircraft - model vector wind
differences. The aircraft by the cursor has a 43
kt wind difference with the model. Uniform
differences between many aircraft and the model
in a particular difference suggest model
problems otherwise, differences suggest aircraft
problems.
32
Distribution of AMDAR data from GSD

• Data are quality-controlled at GSD
• Binary and text data are distributed via GSDs
  MADIS program
• http//madis.noaa.gov/
• Used by many weather service offices
• Used by many research institutions
• Soon to be transferred to operations
• Graphical data available over the web
• http//amdar.noaa.gov/

33
Demonstration of GSDs real-time AMDAR display

• http//amdar.noaa.gov
• Real-time displays are restricted
• JMA has had an account since 2001
• requested by Dr. Masanori OBAYASHI
• but not used recently

34
http//amdar.noaa.gov/java/
35
Zooming in on Japan
36
Can display wind barbs
37
Zooming in on Narita
38
Clicking on an ascent or descent gives a sounding
39
Clicking on Get Text gives the sounding as text
40
A close look at Monday Mornings accident
41
Ascent sounding from aircraft JP9Z4Y55took off
at 2142 UTCNote strong wind direction shear in
lowest levels
42
Higher resolution sounding from aircraft HL7718
(Korean) took off at 2023 UTCNote better
vertical resolution lowest levels
43
Zooming in on the soundingNote 49 kt wind at
1400 ft (AGL)
44
This site is used by weather services and
researchers world-wide

• US NWS
• US FAA
• Contributing US airlines
• US military
• State air quality forecasters
• AMDAR and E-AMDAR management
• Australia, Brazil, Canada, Denmark, Dubai,
  France, Russia, Serbia-Montenegro, So. Africa,
  Spain, Switzerland, others.
• Korean Meteorological Organization has adapted
  our software to make their own displays

45
(No Transcript)
46
(No Transcript)
47
Summary

• AMDAR data improves NWP forecasts
• AMDAR data improves forecasts made by humans
• AMDAR quality monitoring is performed in several
  locations, including GSD
• GSD impact studies show AMDAR is the most
  important data source for many short-term,
  mesoscale forecasts
• AMDAR data are available from GSDs MADIS program
  to approved users
• AMDAR data are available on the web to approved
  users at http//amdar.noaa.gov/
• in plan view
• as soundings

48
Thank you!

• William R. (Bill) Moninger
• NOAA/ESRL/GSD
• R/GSD1
• 325 Broadway
• Boulder, CO 80304
• 303-497-6435
• Bill.Moninger_at_noaa.gov

49
Off-time assimilation

• Traditionally, a model is initialized with RAOBs
  at one on-time (say, 0 UTC)
• and validated with RAOBs at the next on-time 12
  h later.
• The RUC and other modern models can assimilate
  data at off-times
• And generate forecasts to be validated with raobs
  at the next on-time
• (Off-time data consist of much more than AMDAR,
  but well focus on AMDAR)

50
Each cycle gains the benefit of all off-time
observations. There is now enough AMDAR data to
cycle every hour
On
On
Off
Off
Off
Validate with Raob
AMDAR
AMDAR
AMDAR
Raob AMDAR
3-h
6-h
9-h
12-h
9
6
12
0
3
Time (UTC)
51
RUC Wind forecast Accuracy - Sept-Dec 2002
6
9
1
3
12
Analysis truth
RUC is able to use recent obs to improve forecast
skill down to 1-h projection for winds
Verification against RAOB data over RUC
domain RMS vector difference (forecast vs. obs)
this is an important accomplishment -- need to
minimize model disturbances due to imperfect data
(we use DDFI, next slide).
52
RUC Diabatic Digital Filter Initialization (DDFI)
Initial DFI in RUC model at NCEP - 1998 -
adiabatic DFI Diabatic DFI introduced at NCEP -
2006
-30 min -15 min Init 15 min
Backwards integration, no physics
Forward integration, full physics
Obtain initial fields with improved balance
RUC model forecast
53
RUC Diabatic Digital Filter Initialization (DDFI)
Initial DFI in RUC model at NCEP - 1998 -
adiabatic DFI Diabatic DFI introduced at NCEP -
2006
-30 min -15 min Init 15 min
Backwards integration, no physics
Forward integration, full physics
Obtain initial fields with improved balance
Calculate digital-filter-weighted mean of 3-d
fields from each time step over DFI period
RUC model forecast