MODIS Polar Winds in ECMWFs

1
MODIS Polar Winds in ECMWFs Data Assimilation
System Long-term Performance and Recent Case
Studies
Lueder von Bremen, Niels Bormann and Jean-Noël
Thépaut
European Centre for Medium-Range Weather
Forecasts (ECMWF) Reading, United Kingdom
High Latitude NWP Workshop at IARC, Fairbanks,
8-10 October 2003
2
OUTLINE

• History and use of MODIS Polar Winds at ECMWF
• Long-term performance
• a) one satellite (Terra)
• b) two satellites (Terra and Aqua)
• Case study
• Conclusions and prospects

3
History of MODIS AMVs at ECMWF

• trial 3DVAR experiments with MODIS Terra AMVs
  very successful (spring 2001)
• operational archiving of MODIS Terra AMVs since
  July 2002
• 4DVAR experiments with MODIS Terra AMVs
• i) 2 study periods (spring 2001 and summer
  2002, 58 cases)
• ii) operational configuration (T159 60L
  analysis and T511 forecast)
• iii) usage like geostationary AMVs
• iv) modification of mean polar wind analyses
• v) reduction of key analysis errors in case
  study experiment
• vi) good impact on forecast over NH
  (especially over Europe and N. Atlantic)
• operational use of MODIS AMVs (Terra) since 14
  Jan 2003
• operational archiving of MODIS Aqua AMVs since
  Feb 2003
• monitoring since May 2003
• 4DVAR experiments with MODIS Aqua AMVs are
  described here (operational
• analysis and model configuration)

4
Received/Used MODIS Terra AMVs (N. Hemisphere)
IR
Heightlt400hPa
400hPaltHeightlt550hPa
WV Cloud
WV ClEAR
? Need for reduction and quality/bias check
5
Pre-Processing and usage of MODIS AMVs
Usage Land above 400 hPa
Ocean/Ice IR above 700hPa
WV above 550hPa
FG check asymmetric to remove negative OBS-FG
bias (FG-Dep. is scaled with
background error)
Thinning 2 cycles with different box/volume
sizes (96x96km then 140x140km)
PILOT,SONDE
OBS errors
AIREP
AMV
6
Observation Monitoring (WV clear, 400-550 hPa)
OBS-FG OBS-AN
7
LONG-TERM PERFORMANCE
500 hPa Z anomaly correlation (58 days)
Bormann and Thépaut,2003 (spring 2001 and summer
2002)
8
Sensitivity results 4 August 2002, 12 Z
Sensitivity perturbations for streamfunction
around 500 hPa Positive impact of MODIS (negative
differences) occurs where Sensitivity patterns
are large
absolute values
Difference Modis-Control
9
Modis Terra and Aqua data coverage at May 27,
2003 12Z
Light 06Z Dark 12Z
10
Experiments with MODIS (Terra and Aqua)

• 2 study periods Feb. and May 2003, 51 cases
• revision of MODIS impact in general and clean
  control to monitor both
• i) noMODIS (gives chance for clean
  OBS-FG statistic for Terra and Aqua)
• ii) Terra (operational usage (140km
  thinning))
• iii) TerraAqua (140km thinning)
• iv) BOTH200km (Terra and Aqua with 200km
  thinning and QI usage)

11
Mean Polar Wind Analysis/Difference to noMODIS at
400 hPa
noMODIS Terra
TerraAqua
   

m/s
12
RMS of Analysis Increments 400 hPa geopotential
height
noMODIS Terra
TerraAqua
gpdm
13
Used MODIS AMVs (above 400 hPa)

N.Hem.
S.Hem.
Terra
Total Terra
TerraAqua
BOTH200km

• huge amount of Aqua over Antarctica
• Terra and Aqua AMVs not competitive
• moderate increase with BOTH200km only

14
Forecast scores 500 hPa Z anomaly correlation
(51 days)
N.Hem
noModis Terra TerraAqua BOTH200km
S.Hem
(winter and spring 2003)
15
CASE STUDY
Operations noMODIS eSuite eSuite(noMODIS)
TerraAqua half err.cov QI
Day5, 500Z, Europe 12UTC
? Trials with cycles at 30 June 00 and 12 UTC
using noMODIS background
16
500Z Forecast 30 June 2003 12UTC vs Analysis
12h
36h
48h
60h
24h
Analyses
noMODIS
BIASED
17
Study Area 1 Analysis at 30 June 2003, 12UTC
used AMVs 450-650hPa in Exp. BIASED
OBS-Time 0415Z 0555Z 0735Z 0915Z 1050Z
1230Z 1410Z  
18
Analysis at 30 June 2003, 12UTC
vertical velocity at 500 hPa
500Z difference
background
BIASBLACK-BIASED

BIASBLACK
BIASED
Pa/s
19
Short-term forecast at 30 June 2003, 12UTC
vertical velocity at 500 hPa
Operations


3h
Straight model forecast
6h
18h
20
24h forecast error (400Z) at 30 June 2003, 12UTC
BIASBLACK
BIASED
noMODIS
Operations
21
CONCLUSIONS - PROSPECTS

• MODIS Polar winds
• have impact on ECMWFs polar wind analysis
  (stronger over Antarctic)
• introduce analysis increments over very bad
  observed areas
• are consistent with the other sparse wind
  observations
• improve forecast over Europe
• impact is decreased since much more other
  satellite data is in the system