MODIS winds assimilation experiments and impact studies to date at the Met Office Howard Berger, Mar

1
MODIS winds assimilation experiments and impact
studies to date at the Met Office Howard
Berger, Mary Forsythe, Met Office,
Bracknell/Exeter, UK UW-CIMSS Madison, WI
2
MODIS polar winds at the Met Office

• MODIS imagery from Terra and Aqua used to
  generate winds.
• IR(11mm) and WV (6.7mm) channels
• 100 min between overlapping images.
• Time delay of 5-6 hours after valid time before
  winds are available.
• Still experimental. Met Office obtains them via
  ECMWF.

Picture courtesy of CIMSS
3
Two Runs

• Global Forecast Run
• Global Update Run

4
Met Office Operational Schedule
T 0z,6z,12z,18z
T3
T-3
Run Begins at T2
assimilation window
Global Forecast Run

• Assimilate observations valid for 6 hour window
  surrounding valid analysis time
• Only obs valid in window that arrive before T 2
  are used
• Produces 144 hour forecast

5
Met Office Operational Schedule
Global Update Run

• Observations valid for same time as forecast run,
  but given extra time to arrive.
• Produces 6 hour forecast that is used as
  background for successive forecast run.

6
Met Office operational schedule impact on MODIS
wind assimilation
Most MODIS data arrives in time for the update
run.
7
Distribution of winds from Terra and Aqua
AQUA
TERRA
QU00
QU06
Aqua 12,562 Terra 8,661 Total 21,223
Aqua 16,973 Terra 27,535 Total 44,508
QU12
QU18
Aqua 23,206 Terra 21,640 Total 44,846
Aqua 29,229 Terra 7,804 Total 37,033

8
Incentive

• MODIS polar winds provide information on the
  wind field in data sparse regions. This should
  benefit polar wind analyses.
• Investigations at ECMWF with 3DVAR and 4DVAR
  show benefit from assimilating the Terra MODIS
  polar winds.

9
Trial Set-up

• Trial period 12th May - 15th June 2003
• Control low resolution global model run in real
  time
• Experiment Same as control, but use winds from
  Aqua and Terra
• thinned to one wind per 140 km x 140 km x 100 hPa
  box
• blacklisting the following regions
• altitudes below 700 hPa for IR winds over sea
• altitudes below 550 hPa for WV winds over sea
• altitudes below 400 hPa for IR and WV winds over
  land

10
06Z July 03, 2003
11
06Z July 03, 2003
12
Token Model Info Slide

• Low-res Trial Model Characteristics
• Grid-point model (288 E-W x 217 N-S)
• Staggered Arakawa C Grid
• Analysis resolution (216 x 163)
• Approx 100 km horizontal resolution
  (one-half operational resolution)
• 38 vertical levels, hybrid-eta configuration
• Run times 00, 06, 12, 18 Z
• 3-D VAR Data Assimilation

13
Results
14
Anomaly Correlation 500 hPa Geo Height compared
to their own analysis
15
Further verification

• normalized root mean square (rms) error
  against control rms error calculated for
• Mean sea-level pressure (PMSL)
• 500 hPa height (H500)
• 850 hPa wind (W850)
• 250 hPa wind (W250)
• In regions
• Northern Hemisphere (NH)
• Tropics (TR)
• Southern Hemisphere (SH)
• For forecast periods of
• T24, T48, T72 ,T96 , T120

16
Met Office Forecast verification
Verfication against observations
Verfication against analysis against analysis
RMS change (Trial Control) ()
17
Met Office Forecast verification
Verfication against observations
Verfication against analysis against analysis
RMS change (Trial Control) ()
18
Met Office Forecast verification
Verfication against observations
Verfication against analysis against analysis
RMS change (Trial Control) ()
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
Why such different results than ECMWF?

• This study used Terra and Aqua, ECMWF used Terra
• Different season
• Cloud height detection difficult in polar winter
• Inappropriate Observation Errors

23
Further Work
ECMWFs positive results justify further
study! Plan to Complete trial
verification Produce routine O-B
statistics Other tests Retrial

24
Future Trial Options

• Increase Thinning box to 200 km
• Use Quality Indicators for threshold/thinning
• Only use Northern Hemisphere winds
• Modify observation errors
• Change blacklisting criteria
• Different trial season

25
Satellite Winds Superobbing
Hurricane Opal October 1995
Image Courtesy of UW-CIMSS
Howard Berger Mary Forsythe John Eyre Sean Healy
26
Outline

• Background/Problem
• Superob Methodology
• Conclusions/Future Work

27
Problem

• Met Office preliminary impact studies using high
  resolution satellite wind data sets showed
  negative impact (Butterworth and Ingleby, 2000)
• It was suspected that the observation errors were
  spatially correlated, violating an assumption in
  the data assimilation system.
• To account for this negative impact, wind data
  were/are thinned to 2º x 2º x 100 hPa boxes

28

• Bormann et al. (2002) compared wind data to
• co-located radiosondes showing statistically
  significant spatial error correlations up to 800
  km.

Met-7 W V NH Correlations
Correlation
Graphic from Bormann et al.2002
29
Question
Can we lower the data volume to reduce the
correlated error while making some use of the
high-resolution data?
30
Proposed Solution
Average the observation - background
(innovations) within a prescribed 3-d box to
create a superobservation.
31
Advantages

• Data volume is reduced to same resolution that
  resulted from thinning.
• Averaging removes some of the random,
  uncorrelated error within the data.

32
Superobbing Method
33
1) Sort observations into 2º x 2º x 100 hPa
boxes.
28 N
26 N
16 W
18 W
34
2) Within each box Average u and v component
innovations, latitude, longitude and pressures.
28 N
26 N
16 W
18 W
35
3) Find observation that is closest to average
position and add averaged innovation to
the background value at that observation
location.
28 N
26 N
16 W
18 W
36
Superob Observation Error

• Superobbing removes some of the random
  observation error.
• This new error can be approximated by making a
  few assumptions about the errors within the
  background and the observation.

37
Superob Observation Error

• Assume that within a box
• Observation and background errors not correlated
  with each other.
• Background errors fully correlated.
• Background errors have the same magnitude.

38
Superob Observation Error

• Assumptions (cont)
• All of the innovations weighted equally.
• Constant observation error correlation.

39
00z 10 June, 2003. (20 N - 40 N) (0E 30 E)
40
Conclusions
Reducing data volume lowers the effect of
correlated error in satellite winds. Superobbing
does this reduction and reduces the random
error within the observations. The new error
can be approximated by making assumptions about
the structure of observation and background errors
41
Future Work
Model forecast impact studies are
underway Develop quantitative test to evaluate
where superobs will work well and where they will
not.
42
Thanks, Any Questions?