Magdalena D. Anguelova,

1
Validation of satellite-basedestimates of
whitecap coverage Approaches and initial results
16th Conference on Air-Sea Interaction 1115
January 2009, Phoenix, Arizona

• Magdalena D. Anguelova,
• Justin P. Bobak, William E. Asher, David J.
  Dowgiallo,
• Ben I. Moat, Robin W. Pascal, Margaret J. Yelland

Naval Research Laboratory, Washington, DC Applied
Physics Laboratory, University of Washington,
Seattle, WA National Oceanography Centre,
Southampton, UK
2
Long-term goal

• Improve Sea Salt Source Function parameterization
  by
• modeling the high variability of whitecap coverage

or
u wind speed (u10 or u?) ?T atmospheric
stability ( Tair Tsea) X wind fetch d
wind duration Ucur water currents Ts sea
surface temperature S salinity Ck
concentration, type (k) of surface active
materials
3
Framework
Whitecap variability

• Improve existing or develop new models
• Extensive database W various factors
• Measurements W various factors
• Existing W measurements
• Photographs/video images
• Insufficient for extensive database
• Alternative approach From satellites to get
• global coverage
• wide range of meteo environ conditions

The first step
4
Daily map of W
Daily data (swath) for entire 2006, months of
2007 and 2008
5
Satellite-based foam fraction W

• Retrieving W (changes in TB at microwave
  frequencies)
• Using parts of WindSat forward model (v.1.9.6)
• Rough surface emissivity, er
• Atmospheric variables ? atm. correction
• Foam emissivity model, ef

• Independent sources for the input variables
• TB from WindSat
• V, L from SSM/I or TMI
• U10 from QuikSCAT, SSM/I, or GDAS
• Ts from GDAS
• S 34 psu
• Improvements over the published feasibility
  study
• More physical models for er, ef, and atm. corr.
• Independence of the variables
• Minimization of errors.

6
Validation

• Insufficient ground truth values
• Data collection
• Slow and expensive
• Sporadic and non-systematic
• Limited range of conditions
• Fewer in situ-satellite matches in time and
  space
• Different principles of measurement
• Reflectivity in the Visible (photographic/video)
• Emissivity in the Microwave (radiometer)
• Various approaches to circumvent difficulties.

7
Validation approaches

• Historical database of in situ values
• Wind speed formula
• Ship-borne measurements
• Air-borne measurements.
• There are questions and issues
• with each approach.

8
In situ historical data by type
9
10 GHz seems good
10
All Frequencies, H pol
11
Satellite vs Wind formula
March, 2007, 0.5 deg x 0.5 deg

• Wsat more uniform by latitude
• High lat higher W.

Satellite, 18.7 GHz, H pol.
12
Difference maps

• ?W Wsat Wmod

?W 0.041 - ?W 0.44
?W 0.61 - ?W 0.63
10H
18H
13
Polarfront ship data

• Experiment HiWASE on Polarfront ship positioned
  at Station M
• UK colleagues Margaret Yelland, Ben Moat and
  Robin Pascal
• Long-term (Sep 2006 to Sep 2009) measurements of
  W and other variables
• In situ data for W
• Two cameras, daylight restrictions (Mar-Oct)
• Photographic data processed at 3 intensity
  thresholds with AWE technique
• Temporally-averaged values in a time window
  around or close to WindSat pass time
• The effect of time window (minutes to 3 hours)
  was investigated
• WindSat data for W
• Closest pixel to lat/lon position of each in situ
  point
• WindSat low resolution (50 km x 71 km)
• Three frequencies (10, 18, and 37 GHz), H pol.
• The effect of averaging over NºxNº box (e.g.,
  1/2ºx1/2º) was investigated

• In situ data for W
• Two cameras, daylight restrictions (Mar-Oct)
• Photographic data processed at 3 intensity
  thresholds with AWE technique
• Temporally-averaged values in a time window
  around or close to WindSat pass time
• The effect of time window (minutes to 3 hours)
  was investigated
• WindSat data for W
• Closest pixel to lat/lon position of each in situ
  point
• WindSat low resolution (50 km x 71 km)
• Three frequencies (10, 18, and 37 GHz), H pol.
• The effect of averaging over NºxNº box (e.g.,
  1/2ºx1/2º) was investigated

14
In situ historical data by type
15
Polarfront to historical in situ
16
WindSat matched to Polarfront
17
In situ and satellite winds

• 180-min time window

18
In situ and satellite winds

• 180-min time window

19
RASSI Experiment

• RAdiometry and Sea Surface Imagery (2007)
• North Atlantic, Gulf of Mexico (Hurricane Dean)
• High altitude 6.7 km (20,000 ft)
• Clear sky to partial cloud cover
• Radiometric measurements
• APMIR (Airborne Polarimetric Microwave Imaging
  Radiometer)
• Channels available (GHz) 37VH34, 19VH34, 6.6VH,
  6.8VH, 7.2VH (some data at channels at 10.7 and
  22.235)
• Footprint roughly 1x2 km from on 19 and 37 GHz
• Video measurements
• High resolution video camera
• Field of view of 159 m by 119 m.

20
RASSI Experiment

• RAdiometry and Sea Surface Imagery (2007)
• North Atlantic, Gulf of Mexico (Hurricane Dean)
• High altitude 6.7 km (20,000 ft)
• Clear sky to partial cloud cover
• Radiometric measurements
• NRLs APMIR (Airborne Polarimetric Microwave
  Imaging Radiometer)
• Channels available (GHz) 37VH34, 19VH34, 6.6VH,
  6.8VH, 7.2VH (some data at channels at 10.7 and
  22.235)
• Footprint roughly 1x2 km from on 19 and 37 GHz
• Video measurements
• High resolution video camera
• Field of view of 159 m by 119 m.

21
RASSI Experiment

• RAdiometry and Sea Surface Imagery (2007)
• North Atlantic, Gulf of Mexico (Hurricane Dean)
• High altitude 6.7 km (20,000 ft)
• Clear sky to partial cloud cover
• Radiometric measurements
• NRLs APMIR (Airborne Polarimetric Microwave
  Imaging Radiometer)
• Channels available (GHz) 37VH34, 19VH34, 6.6VH,
  6.8VH, 7.2VH (some data at channels at 10.7 and
  22.235)
• Footprint roughly 1x2 km from on 19 and 37 GHz
• Video measurements
• UWs High resolution video camera
• Field of view of 159 m by 119 m.

22
Hurricane Dean flight
23
In situ historical data by type
24
RASSI foam vs historical in situ
25
RASSI foam vs historical in situ
26
RASSI vs WindSat pairs
Wind dependent AB factor 11 to 15 using Monahan
Woolf (1989) parameterizations for AB and A
27
RASSI vs WindSat pairs
Wind dependent AB factor 11 to 15 using Monahan
Woolf (1989) parameterizations for AB and A
28
Summary

• Compensate using different approaches
• Historical in situ data
• Wind speed formula
• Direct validation with ship-borne photographic
  data
• Direct validation with air-borne video data
• Results
• Ball-park in magnitude compared to in situ data
• More uniform latitudinally than wind formula
• Direct validation shows
• Wsat overestimate at low winds and
• Relatively good estimate at high winds
• Future work
• More match-ups of in situ and satellite data
• Indirect validation (with other variables, not
  directly W)
• Tuning of the satellite-based algorithm.

• Difficulties in validating satellite-based foam
  fraction
• Amount of data
• Conditions covered
• Principle of measurements
• Compensate using different approaches
• Historical in situ data
• Wind speed formula
• Direct validation with APMIR/video data
• Direct validation with ship data
• Results
• Ball-park in magnitude compared to in situ data
• More uniform latitudinally than wind formula
• Direct validation shows
• underestimate at low winds and
• over estimate at high winds
• How much of this result is correct?
• Future work
• More match-ups of in situ and satellite data
• Indirect validation (with other variables, not
  directly W)

?
29
Additional slides
30
Match-up issues

• Only 4 full swaths
• from GDAS,
• Large temporal mismatch for other GDAS match-ups
• Chunks of swaths for most passes
• due to QuikSCAT passes CROSSING the WindSAT
  passes
• Reflects on the number of samples available for
  low and high latitudes.

31
Samples available for 1 month

• High latitudes with high winds are under
  represented.

32
Frequency and polarization dependence

• Understanding the info each freq and pol gives
  for W
• Research foam skin depth
• The effect of foam thickness on the skin depth
  and foam emissivity
• The results could be important for gas exchange
  (CO2 and other gases)
• Combining suitable freqs and pols in one
  oceanographically representative W

33
Compare to in situ AB data
34
Low freqs close to in situ AB
35
18 GHz

• Very similar to 23 GHz
• Atmosphere influence?

36
All frequencies, V pol
37
All Frequencies, H pol
38
Biases

• Pair binned data
• ?W Wsat-Wins
• Plot bins with high count.

Freq (GHz), H pol. 6 10 18 23 37
39
Satellite vs Wind formula
March, 2007 0.5 deg x 0.5 deg
Wind speed formula
Satellite, 10.7 GHz, H pol.
40
Ship cruises

• Two ship cruises
• in 2006 and 2007
• Ian Brooks and Margaret Yelland, UK
• Data for
• Foam fraction direct validation
• Sea-salt aerosol flux indirect val.
• Matchups with WindSat data
• Spatial and temporal
• Video data availability for about 10 points
• Cruise data still processed

41
In situ and satellite winds (raw)
42
Binned by wind speed
43
Compare in situ satellite pairs
44
Compare in situ satellite pairs
45
RASSI measuring configuration
46
Wind speed conditions
47
Wind vector field during the flight