MODIS Winds Assimilation Impact Study with the CMC Operational Forecast System R

1
MODIS Winds Assimilation Impact Studywith the
CMC Operational Forecast SystemRéal
SarrazinData Assimilation and Quality
ControlCanadian Meteorological
CentreMeteorological Service of Canada
Workshop on Short-to-Medium Range Regional NWP in
the Arctic and Antarctic Fairbanks, Alaska,
October 8-10, 2003
Environnement Canada Environment
Canada Centre météorologique canadien
Canadian Meteorological Centre
2
(No Transcript)
3
Observations
4
Satellite Winds Selection Procedure
Geostationary Satellite winds GOES-P / W / E,
METEOSAT-7 / 5

• Time window within 90 minutes from analysis
  time
• Levels VI below 700 hPa, WV above 400 hPa, IR
  all levels
• Wind speed gt 2.5 m/s
• Angle lt 55 deg.
• Land Mask over ocean, over land south of 20N
  and above 400 hPa
• quality indicator above threshold value
• METEOSAT QI gt 85,
• GOES-W / E RFF
  lt400 401-700 gt700 hPa
• 
  extra-tropics 65 70
  75
• tropics
  70 75 80
• horizontal thinning 1.5 X 1.5 deg.
  (priority obs time, QI)

Quality Control

• Background check done before the horizontal
  thinning
• during the analysis, Variational QC with
  asymetric condition for the AMVs

5
No SATWINDS experiments, 17 June 2002 to 31 July
2002 RMS of forecast Wind speed errors at 250 hPa
e02cntrl control, e02nosw no AMVs, e02noto
no TOVS, e02nohu no HUMSAT, e02nosat no
satellites
6
No SATWINDS experiments, 17 June 2002 to 31 July
2002 anomaly correlation GZ 500 hPa
7
MODIS Winds Assimilation Impact Trial
Data obtained by ftp from CIMSS in near real
time Assimilation Period 18 July 2003 to
23 August 2003,
same cut-off time as for the operational
observations, T6 at 06/18UTC, T9 at
00/12UTC Evaluation Period 5 weeks from
20 July to 23 August 2003, 6-day
Forecast twice per day at 00 and 12 UTC,
from the analyses of the assimialtion cycle.
verification scores against radiosonde
observations and against analyses.
8
RFF quality indicator versus observation minus
first guess statistics for High level MODIS
winds, infrared channel (using the control
first-guess). Plotted are the RMSVD, average wind
speed, wind speed bias and number of observations
per 0.01 bin.
Arctic above, Antarctic below
Period 20 July 08 August 2003

• The average model wind speed is slightly higher
  in the Arctic
• But the RMSVD values are lower than the
  Antarctic
• A NRMSVD gives higher values for the Antarctic
• WV winds (not shown) exhibit similar
  characteristics

9
IR channel, statistics stratified in 3 layers,
Arctic Region
The characteristics of the statistics are similar
to those of GOES winds Including the shift of the
distribution toward higher RFF values (and lower
mean wind speed) for lower levels winds
10
QI quality indicator versus observation minus
first guess statistics for High level MODIS
winds, infrared channel (using the control
first-guess). Plotted are the RMSVD, average wind
speed, wind speed bias and number of observations
per 0.01 bin.
Arctic above, Antarctic below, Period 20 July
08 August 2003

• RMSVD values are almost constant, values for the
  Antarctic are higher
• Average wind speed increases with increasing QI
  values so NRMSVD increases

11
Satellite Winds Selection Procedure
MODIS winds Terra / Aqua

• Time window within 90 minutes from analysis
  time
• Levels IR above 700, WV above 550 hPa
• Wind speed gt 2.5 m/s
• Land Mask over ocean, over land above 400 hPa
• quality indicator above threshold value RFF
  lt400 401-700
• 
  65
  70
• horizontal density thinning average of 180 km
  (priority obs time, qi)

Quality Control

• Background check done before the horizontal
  thinning
• during the analysis, Variational QC with
  asymetric condition for the AMVs
• (no observation height reassignment)

12
Example of AMVs distribution for one analysis, 28
august 12UTC
13
MODIS Winds trial, period 20 July 2003 to 23
August 2003 24-hour Forecasts Verification
against radiosondes, N of 60N, Arctic
RMS solid lines Bias dashed lines Control
blue lines MODIS red lines There is a small
negative impact (increased rms) on the errors of
the forecasts in the Arctic UU east-west wind
component UV wind speed GZ geopotential
heights TT temperature
14
MODIS Winds trial, period 20 July 2003 to 23
August 2003 24-hour Forecasts Verification
against radiosondes, S of 60S, Antarctic
RMS solid lines Bias dashed lines Control
blue lines MODIS red lines There is a
negative impact on the errors of the forecasts in
the Antarctic UU east-west wind
component UV wind speed GZ geopotential
heights TT temperature
15
MODIS Winds trial, period 20 July 2003 to 23
August 2003 24-hour Forecasts Verification
against radiosondes, Northern Extratropics
RMS solid lines Bias dashed lines Control
blue lines MODIS red lines There is little
impact on the errors of the forecasts in the
Northern Hemisphere UU east-west wind
component UV wind speed GZ geopotential
heights TT temperature
16
MODIS Winds trial, period 20 July 2003 to 23
August 2003 24-hour Forecasts Verification
against radiosondes, Southern Extratropics
RMS solid lines Bias dashed lines Control
blue lines MODIS red lines There is a small
negative impact on the errors of the forecasts in
the Southern Hemisphere UU east-west wind
component UV wind speed GZ geopotential
heights TT temperature
17
MODIS Winds trial, period 20 July 2003 to 23
August 2003 Forecasts Verification against
analyses, Anomaly correlation, N of 60N and S of
60S
18
MODIS Winds trial, period 20 July 2003 to 23
August 2003 Forecasts Verification against
analyses, Anomaly correlation, extratropics
19
MODIS Winds trial, period 20 July 2003 to 23
August 2003 Forecasts Verification against
analyses, Wind Speed errors RMS, N of 60N and S
of 60S
20
Conclusion
For this first relatively short
trial, Verifications of the forecasts against
radiosondes show a small negative impact from the
MODIS winds, especially for the
Antarctic Verifications of the forecasts against
analyses show mixed results generally negative
but with some positive impacts on winds speeds
forecast quality at mid levels Longer trials are
necessary before implementation