Meteorological Impacts and Benefits of AMDAR Data
Lee CronceRalph Petersen
Cooperative Institute for Meteorological Satellite Studies (CIMSS)Space Science and Engineering Center (SSEC)
University of Wisconsin - Madison
AMDAR Regional Science and Technology WorkshopMexico City, Mexico9 November 2011
Atmospheric Data Realities
• Vertical variation of temperature, moisture and wind within the atmosphere is what drives Atmospheric Science
• In situ profiles of these data (i.e., radiosondes) remain the backbone of any NWP analysis system
• The details within these profiles are especially important for recognizing and predicting hazardous weather events
Atmospheric Data Realities
• Vertical variation of temperature, moisture and wind within the atmosphere is what drives Atmospheric Science
• In situ profiles of these data (i.e., radiosondes) remain the backbone of any NWP analysis system
• The details within these profiles are especially important for recognizing and predicting hazardous weather events
• The locations of radiosonde sites are sparse, and the number of radiosonde reports is decreasing worldwide
Atmospheric Data Realities
• Vertical variation of temperature, moisture and wind within the atmosphere is what drives Atmospheric Science
• In situ profiles of these data (i.e., radiosondes) remain the backbone of any NWP analysis system
• The details within these profiles are especially important for recognizing and predicting hazardous weather events
• The locations of radiosonde sites are sparse, and the number of radiosonde reports is decreasing worldwide
• Satellites provide global coverage; however, not at detail necessary (especially near the surface)
• AMDAR fills this void!
AMDAR in a Nutshell• Temperature and Wind Observations from
Commercial Aircraft– High quality, high resolution data– Available at:• Flight Level• In Ascent / Descent
– Instruments already on aircraft– Economical (~100 times less expensive than radiosondes)
– Asynoptic (not only available at 00 and 12UTC, not a problem for NWP)
– Looks, feels, tastes like radiosonde data– Retrieved through ACARS and MDCRS
Usage of AMDAR Data
WHO WOULD BENEFIT FROM THIS
DATA SOURCE?
How does NCEP use AMDAR data?In all of its atmospheric models
• NCEP runs a suite of Atmospheric and Oceanic models to meet a variety of user needs. AMDAR data are used in:– Climate (Coupling Atmosphere and Ocean)– Global (Medium Range Forecasts)• 4/day - Deterministic and Probabilistic
– Mesoscale (Higher-Resolution Weather Forecasts)• 4/day - Deterministic and Probabilistic• Rapid Update Cycle (RUC) [soon to be Rapid Refresh Model]
– Hourly - Aviation and Hazardous Weather» Mexico included in coverage, so immediate use available
• Vertical profiles of wind, temperature and humidity are the foundation of every NWP system
• NCEP has been using AMDAR data in its NWP models for over 10 years
• Over 300,000 reports arrive daily– Data delivered in real-time 24 hours daily– Most contain wind and temperature only
• Increasing numbers include humidity– The data arrive in BUFR format
• The program is a cooperative venture between data providers and users– Everyone benefits from the results
How does NCEP use AMDAR data?In all of its atmospheric models
“Rule of Thumb”• In Numerical Weather Prediction (NWP), one bad
observation does more damage than the benefit that comes from 100 good observations!
• AMDAR data are extremely accurate and reliable, but
– Good Quality Control of all observations is essential• Requires multiple observations
How does NCEP use AMDAR data?In all of its atmospheric models
A major advantage of AMDAR data – multiple observations corroborate each otherWeekly Data counts by Cycle
Data Volume/Coverage by Layers
700-300mb
300-100mb
1000-700mb
Six hours of data
Note locations Of Ascent/descent Reports
←←
FORECAST IMPACTS OF AMDAR DATA
Determining Forecast Improvement from increased AMDAR volume – Use Wind forecasts as a measure of impact
•During weekday, when more AMDAR reports are available, short range forecasts are consistently better
•0000-1200 UTC (overnight) AMDAR volume average• Tu-Sa >70,000 reports• Su-Mo only ~25,000 reports
•Difference is primarily due to lack of parcel delivery flights
General Observation
Quantifying these Observations using theRapid Update Cycle - RUC
RUC is designed to produce hourly analyses andupdates to very short range forecasts (0-12 hrs)
Real-time 1-hourly analysis/forecast cycle Analyses intended to fit data very closely Forecasts only from 3 to 12 hours into future In general, 3 hr RUC wind forecasts are more accurate than 12 hr forecasts
Examination of verification against Radiosonde observations
Weekend minus Weekday 3 hr
Wind Forecast Errors for Jan-Oct
RUC Wind forecasts-Verification against raob data
Weekend-weekday 3h wind fcst error
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
850 700 500 400 300 250 200 150
Pressure level (hPa)
Win
d fc
st e
rror
dif
f (m
/s)
0.35 m/s / ~5.0 m/s= 7% better forecasts during weekdays due to more AMDAR reports at 200 hPa
Off-time data on weekends produces
less impact, especially after
reduced overnightpackage carriers reports
Hourly AMDAR VolumeReceived at FSL (ESRL)
2-15 Sept 01(starting 00z 2 Sept)
0
1000
2000
3000
4000
5000
6000
1 25 49 73 97 121 145
Hour of week
Air
craf
t rep
orts
/ ho
ur
Series1
0
1000
2000
3000
4000
5000
6000
1 25 49 73 97 121 145
Hour of week
Air
craf
t rep
orts
/ ho
ur
Series1
Su Mo Tu We Th Fr Sa
2-8 Sept 01
9-16 Sept 01
Su Mo Tu We Th Fr Sa
Notable reductions of aircraft dataavailable to RUC at FSL
on weekends andimmediately after Sept. 11, 2001
Improvement in 3 hr over 12 hr wind forecasts
during September 2001
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
1 5 9 13 17 21 25 29 33 37 41
12h periods - 1 Sept 01 to 21 Sept
12h-
3h fo
reca
st e
rror
diff
Series1
RUC 250mb wind forecasts verified against raob data
Period of data outage11-13 Sept 2001
Forecasts fromoperational RUCrun at NCEP
11-13 September 2011• No AMDAR data• 20% loss of 3hr
RUC wind forecast skill at 250mb
• 3 hr fcst skill ≅12hr skill
• No skill added by other off-time reports!!!
Impact of AMDAR Ascent/Descent Data in Rapid Update Cycle (RUC) forecasts
• Test performed using operational 20 km RUC– Ran data assimilation / forecast system for 3 weeks in June
2002 using two configurations:1. Including all data2. Eliminated aircraft data below 350 hPa
– Kept High-level En-route Data– Ignored Ascent /Descent Data
– Compared analyses and all forecasts (3, 6, 9, 12 Hr) against radiosonde at 00 and 12 UTC over CONUS
– Results expressed in improvement due to Ascent/Descent Data
EMC OSE by Ralph Petersen, Geoff Manikin and Dennis Keyser
Impact of AMDAR Ascent/Descent Data in Rapid Update Cycle (RUC) forecasts
• Question 1– What was the effect of the addition of
ascent/descent data on the data assimilation system and resulting 00 and 12 UTC analyses?
•Significant improvement by including Ascent / Descent data• Positive effects at all levels• Greatest effect at 30,000’ and below• Positive impact on Winds, Temp and RH.
Normalize error: compares
forecast differenceswith
overall forecast error
Impact of AMDAR Ascent/Descent Data in Rapid Update Cycle (RUC) forecasts
• Question 2– What was the effect of these analysis differences
on the 12 hr forecasts?
Tropospheric Improvementsup to twice
those in changingRUC from 40 to 20 km
•Significant improvement by including Ascent / Descent data• Positive effects at all levels on Winds, Temp and RH
• Above 25,000’, impact comparable to analysis differences• Below 25,000’, impact still large - but slightly smaller than in analysis
Impact of AMDAR Ascent/Descent data in Rapid Update Cycle (RUC) forecasts
• The fundamental purpose of the RUC is to use ‘off-time’ data to make repeated corrections to traditional ‘on-time’ model guidance
• Question 3– How did the continued data assimilation affect
model performance?
•After 9 hrs of continued use of ascent/descent data, tropospheric forecasts have improved by yet another 1-2%
Tropospheric Improvementsare 2-3 times greater than
those in changingRUC from 40 to 20 km
12 Hr Forecast Error – Red 3 Hr Forecast Error – Blue
Both forecasts valid at same times
Impact of AMDAR Ascent/Descent data in updating operational RUC forecasts
10–20% improvementat all levels
from forecast updates
Descent
Impact of AMDAR Ascent/Descent Data in Rapid Update Cycle (RUC) forecasts
• The fundamental purpose of the RUC is to use ‘off-time’ data to make repeated corrections to traditional ‘on-time’ model guidance
• Overall question now becomes:
– How much of the impact was the result of including ‘off-time’ ascent / descent data?
Assimilation/forecasts with Ascent/Descent Data– RedAssimilation/forecasts without Ascent/Descent Data– Blue
Difference between 12 hr operational RUC forecast and a later 3 hr forecast (valid at the same time but using additional asynoptic
reports) from systems with & without ascent/descents
Lack ofascent/descent data
in assim./fcstseliminates virtually all tropospheric benefits
of off-time updates anddegrades upper-levels
Descent
LOCAL APPLICATIONS OF
AMDAR DATA
• Severe Weather– Capping Inversions– Convective Instability– Wind Shear
• Precipitation and Type– Timing, Location,
Intensity
• Fog Onset/Dissipation– Trapping Inversion
Development/Decay– Calm Winds
• Air Quality/Fire Weather– Wind, Stability, Mixing,
Extended Coverage
Local Applications
Low-Level Wind Shear
• Based on this observation, the aviation forecaster was able to update the TAF and begin the LLWS more than 3 hours earlier than the prior forecast.
• Green Bay, Wisconsin, 29 October 2005• LLWS was forecast to begin after 0600 UTC in the TAF• Aircraft soundings near 0120 UTC already showed LLWS
Low Ceilings, Visibilities and Fog• Detroit, Michigan, 4 February 2005• Soundings near 2230 UTC showed light boundary layer winds, near-
surface moisture, dryness above• Commonly favorable conditions for fog development
• Based on the observations, the TAFs for 09 and 12 UTC were amended, reducing visibilities to ½ mile.• METARS showed that visibilities did decrease
KDTW 0532z 00000kt 2sm br clr KDTW 0739z 17003kt 1 3/4sm br r04/ 1000v3500 KDTW 0936z 17004kt 1/4sm fg r04/ 0500v0600 KDTW 1154z 16004kt 1/4sm fg r04/ 2800v0600
• Buffalo, New York, 15 December 2005• Forecasters initially were calling for larger snow accumulations• AMDAR temperature profile shows a larger than expected warm layer aloft
Precipitation Type
• With the existence of this deep warm layer aloft, forecasters amended the forecast calling for smaller snow accumulations and increased chances for sleet and freezing rain
Convective Storms• Central Wisconsin, 6 July 2005• Linear mesoscale convective system expected to persist into Wisconsin• Severe thunderstorm watch was issued at 1530 UTC for most of Central
Wisconsin
Convective Storms• Aircraft soundings from watch area at watch issuance and later showed
strong capping inversion unlikely to break• Forecasters lowered the chance for storms and the severe
thunderstorm watch was cancelled
• Very dry air could be seen on aircraft soundings earlier in the day when the Red Flag Warning was issued
• Later soundings showed there was sufficient dry air in other parts of the forecast area to expand the warning• Temperature >75F, RH <25%, winds >25 mph
Fire Weather• Northern and Central Wisconsin, 15 June 2006• Aircraft data showed extremely dry conditions coupled with the
potential for high winds due to mixing
In Summary
• AMDAR is a very important and necessary data set
• Fills spatial and temporal voids apparent in radiosonde and satellite data sets
• En-route data, but more so, ascent/descent data are vital to NWP skill
• Not just a NWP benefit, but an important local forecast area data set