Data assimilation of polar orbiting satellites at ECMWF
description
Transcript of Data assimilation of polar orbiting satellites at ECMWF
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Data assimilation of polar orbiting satellites at
ECMWF
Tony McNally
ECMWF
ECMWFUse of Satellite Data at ECMWF – Tony McNally
1. Data assimilation
2. Radiance observations from polar orbiting satellites
3. scientific challenges
Overview
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Numerical Weather
Prediction (NWP)
Historical reanalysis for
climate research
Environmentalmonitoring and
modelling
Main areas of activity at ECMWF
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Numerical Weather
Prediction (NWP)
Deterministic Monthly Seasonal
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Environmentalmonitoring and
modelling
Estimating greenhouse gas concentration and
flux inversion
Monitoring and forecasting trajectory
of dust events
Monitoring and forecasting trajectory
of volcanic events
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Re-analysis for climate
research
Trend analysis of climate parameters
Improved climatology for process studies
Cleansed historical observation data sets
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Numerical Weather
Prediction (NWP)
Historical reanalysis for
climate research
Environmentalmonitoring and
modelling
ECMWFUse of Satellite Data at ECMWF – Tony McNally
DATA ASSIMILATION
Numerical Weather
Prediction (NWP)
Historical reanalysis for
climate research
Environmentalmonitoring and
modelling
ECMWFUse of Satellite Data at ECMWF – Tony McNally
What is data assimilation ?…in essence data assimilation is the combination of information from a model and observations to produce a best estimate of the state of the atmosphere (the analysis) ….
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Forecast model
Observations
Assimilation algorithm
Super-computer
Key elements of the assimilation system:
])H[(])H[(
)()()(1
1
xyxy
xxxxxJT
bT
b
R
B
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The forecast model
Xt=0 Xt=t
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The forecast model
91 vertical levels from the surface to 0.01hPa (approx: 80Km)
Global T1279 spectral resolution(16km grid point spacing)
Physical and dynamical processes updated every 10 minutes
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The forecast model
91 vertical levels from the surface to 0.01hPa (approx: 80Km)
Global T1279 spectral resolution(16km grid point spacing)
6,300,000,000,000,000 floating point operations
for a single 10 day forecast
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The Observations
Yobs
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Operational Global Observing Network
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Operational Global Observing Network
~ 60,000,000 observations used every 12 hours
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The Algorithm
4D-Var
(four dimensional variational analysis)
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The 4D-Var Algorithm Jb
])H[(])H[(
)()()(1
1
xyxy
xxxxxJT
bT
b
R
B
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The 4D-Var Algorithm Jo
])H[(])H[(
)()()(1
1
xyxy
xxxxxJT
bT
b
R
B
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The Super-computer
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Super computer configuration
June 2010
ECMWFUse of Satellite Data at ECMWF – Tony McNally
The assimilation of polar satellite observations
ECMWFUse of Satellite Data at ECMWF – Tony McNally
On NOAA / NASA / EUMETSAT polar orbiting spacecraftHigh resolution IR Sounder (HIRS), Advanced Microwave Sounding Unit (AMSU), Atmospheric IR Sounder (AIRS), Infrared Atmospheric Sounding Interferometer (IASI), Advanced Microwave Scanning Radiometer (AMSR), TRMM (TMI), Cross-track Infrared Sounder (CrIS)
On DMSP polar orbiting spacecraftSpecial Sensor Microwave Imager (SSMI,SSMI/S)
Note: the vast majority of data comes from near-nadir passive sounders
Some of the most important satellite instruments for NWP…
ECMWFUse of Satellite Data at ECMWF – Tony McNally
8461 infra-red radiances measured by the IASI instrument
Example of a modern satellite sounding instrument… IASI
ECMWFUse of Satellite Data at ECMWF – Tony McNally
What benefits do polar satellite observations bring
to NWP ?
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Evolution of ECMWF NWP forecast skill
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Evolution of NWP forecast skill
1987*
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Anomaly correlationgeopotential height
500 hPa
Anomaly correlationgeopotential height
500 hPa
Forecast skill without polar satellites ?
S.H.: ~3 days at day 5
N.H.: ~2/3 to 3/4 of a day at day 5
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Snowfall forecasts over North Eastern USA, 3 days in advance of the 19th December 2009 at 12z. The assimilation system with NO POLAR SATELLITES fails to predict the snow storm that caused widespread disruption to the US east coast. Contours start at 5cm and are at 5cm intervals. Red indicates more than 20cm.
NO POLAR
ECMWF OPS VERIFICATION
Forecast skill without polar satellites ?
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Forecasts of Mean Sea Level Pressure, 5 days in advance of the 30th October 2012 for the landfall of Hurricane Sandy. Forecasts from an assimilation system with no polar satellites fails to predict the correct landfall of the storm that caused widespread damage and loss of life to the US east coast.
NO POLAR SATECMWF OPS VERIFICATION
Forecast skill without polar satellites ?
5 day forecast: Base time 2012-10-25-00z Valid Time: 2012-10-30-00z
ECMWFUse of Satellite Data at ECMWF – Tony McNally
What challenges do polar satellite observations
present ?
ECMWFUse of Satellite Data at ECMWF – Tony McNally
They DO NOT measure TEMPERATUREThey DO NOT measure HUMIDITY or OZONEThey DO NOT measure WIND
…these instruments measure the radiance L that reaches the top of the atmosphere at given frequency v …
…ECMWF assimilates these radiances directly(not retrievals of temperature, humidity etc…)
What do these instruments measure ?
ECMWFUse of Satellite Data at ECMWF – Tony McNally
dzdz
dzTBL
0
)())(,()(
+ Surfaceemission
+Surface
reflection/scattering
+ Cloud/raincontribution
+ ...
Planck source term* depending on temperature of the atmosphere
Absorption in theatmosphere
Other contributions to themeasured radiances
Our description of the atmospheremeasured by the satellite
The radiative transfer equation
])H[(])H[(
)()()(1
1
xyxy
xxxxxJT
bT
b
R
BThe RT equation is part of the 4DVar operator that maps the model state X vector into the observation space Y
ECMWFUse of Satellite Data at ECMWF – Tony McNally
1. Limited vertical resolution
2. Sensitivity to cloud and rain
3. Systematic error
Specific Science Challenges
ECMWFUse of Satellite Data at ECMWF – Tony McNally
1. Limited vertical resolution
ECMWFUse of Satellite Data at ECMWF – Tony McNally
o1 2
Ab
so
rpti
on
Frequency
Transmission Weighting function
Pre
ss
ure
1. Limited vertical resolution
dzdz
dzTBL
0
)())(,()(
ECMWFUse of Satellite Data at ECMWF – Tony McNally
AMSUA 15 channels IASI 8461 channels
1. Limited vertical resolution
ECMWFUse of Satellite Data at ECMWF – Tony McNally
1. Limited vertical resolutionIf we consider the assimilation of these radiances as correcting errors in the background state, the success will depend crucially on the size and vertical structure of the background errors (EDA / EnKF etc…)
ECMWFUse of Satellite Data at ECMWF – Tony McNally
2. Sensitivity to cloud and rain
ECMWFUse of Satellite Data at ECMWF – Tony McNally
dzdz
dzTBL
0
)())(,()(
+ Surfaceemission
+Surface
reflection/scattering
+ Cloud/raincontribution
+ ...
Our description of the atmospheremeasured by the satellite
2. Sensitivity to cloud and rain
The cloud uncertainty in radiance terms may be an order of magnitude larger than the T and Q signal (i.e. 10s of kelvin compared to 0.1s of Kelvin!
ECMWFUse of Satellite Data at ECMWF – Tony McNally
surfacesurface
full cloud at 500hPa
dR/dT500 = 0
dR/dT* = 1
dR/dT500 = 1
dR/dT* = 0
Weighting function non-linearity
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Sensitive areas and cloud cover
Location of sensitive regions
Summer-2001(no clouds)
monthly mean high cloud cover
monthly mean low cloud cover
sensitivity surviving high cloud cover
sensitivity surviving low cloud cover
From McNally (2002) QJRMS 128
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Cloud obscured singular vector ?
Some extra overcast observations are used – leading to some possibly important analysis differences in a sensitive area …
500hPa analysis difference (K)
Forecast impact from cloudy data!
ECMWFUse of Satellite Data at ECMWF – Tony McNally
3. Systematic error
(global influence)
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Globally averaged bias correction estimates for MSU channel 2
Warm-target temperatures for MSU on NOAA-14
3. Systematic error … data
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Shifts in upper-stratospheric temperature reanalysis
The transition from SSU Ch3 to AMSU-A Ch14 is clearly visible in global mean temperatures at 5hPa and above
The use of weak-constraint 4D-Var can (only) partially address this problem
This problem cannot be completely solved unless the forecast model is free of bias
ERA-Interim
Global mean temperature anomalies in the upper stratosphere
ERA-40
JRA-25
NCEP
3. Systematic error … model
ECMWFUse of Satellite Data at ECMWF – Tony McNally
Response to Pinatubo: HIRS Ch11 bias corrections
Volcanic aerosols in the lower stratosphere:
• Cooling effect on radiances • Not represented in radiative transfer model• ERA-Interim: Change the bias correction • ERA-40: Change the humidity increments
Bias corrections for HIRS Ch11 (tropical averages)
Bias corrections for NOAA-12:
• In ERA-Interim, correct initialisation followed by a gradual recovery • In ERA-40, bias held fixed
3.Systematic error..atmosphere
ECMWFUse of Satellite Data at ECMWF – Tony McNally
1. Data assimilation lies at the centre of NWP, climate re-analysis and environmental monitoring
2. Radiance observations from polar orbiting satellites are the single most influential component of the global observing system
3. Great progress has been made, but significant scientific challenges remain to advance the use of these observations
Summary
ECMWFUse of Satellite Data at ECMWF – Tony McNally
)(][][ 1b
TTba xyxx HRHBHHB
])H[(])H[()()()( 11 xyxyxxxxxJ Tb
Tb RB
)(][][ 1b
TTba xyxx HRHBHHB
Cost function:
Solution:
Sa = B - HB
Solution error covariance:
The 4D-Var Algorithm
Correction term