README: UH Motion Corrected Sonic Wind RevB,  BWB, Aug 2014

**************************************************************************************

These files are 10 Hz u,v,w sonic wind and sonic temperature data for the 2013
HiWinGS cruise.  This is the RevB data set, updating the motion corrections from
the RevA dataset.  The sonic is a Gill model R2.  Motion corrections are
derived from 6 channel acceleration and angular rate data (Systron-Donner MotionPak).

Timestamps are precisely 10 Hz, synchronized with GPS UTC time, and free of clock drift
and missing values.

MEASUREMENT HEIGHT IS 16.29 M ABOVE THE WATERLINE ON THE BOW IMET TOWER.

Hourly wind data files contain the following variables:

	TimeStampStr: 	Date/Time UTC as string:  MM/DD/YY HH:MM:SS.sss
	Time_GPS_sec: 	Time in seconds from 0000 hrs, 1/1/2013.
	U: 				Relative wind velocity along the streamline.
	V: 				Cross-stream wind velocity (hourly mean = 0).
	W: 				Vertical wind velocity (hourly mean = 0).
	RelWdir: 		Relative wind direction w/respect to ship's bow. Starboard positive.
	SonicT_C: 		Sonic temperature computed from Gill R2.
	
The Gill R2 does not output sonic temperature directly - only speed of sound.
SonicT_C was computed from speed of sound with the NOAA procedure as follows (MATLAB):

    %compute sonic temperature, NOAA R2 formula
    SpdSnd = Winds.Spd_Snd'.*16 + 290;  % speed of sound in m/s
    sonT = (SpdSnd.*SpdSnd + (u.*u + W.*W)/2. + v.*v)/403-273.15;

Motion corrections follow the Edson98 scheme, with the exception that ship's heading
(gyro) is set to zero and the resulting winds are ship-relative, i.e. not in earth
coordinates.

In RevB the calibration of UH accelerometer channels was adjusted by
~5-8% following comparison with the NOAA/PSD MotionPak, which has a more recent
calibration and is judged more accurate.

As an additional step, decorrelation with respect to platform acceleration and
velocity was applied to 10Hz winds after the Edson correction but prior to
rotating into the streamlines (MATLAB):

    % Sum platform velocities and wind to get motion corrected winds
    uvw = uvw + uvwplat;	%final step in Edson98 correction
    
    % Decorrelate wind and motion data. Sequentially remove correlation
    % between wind and platform acceleration/velocity for each axis.  Cross
    % correlation coefficients computed as in Edson 2011 GASEX paper.
    
    Cua = cov(detrend(uvw(1,:)),detrend(acc(1,:)));
    muua = Cua(1,2)/Cua(2,2);
    uvw(1,:) = uvw(1,:) - muua*detrend(acc(1,:));
    Cuv = cov(detrend(uvw(1,:)),detrend(uvwplat(1,:)));
    muuv = Cuv(1,2)/Cuv(2,2);
    uvw(1,:) = uvw(1,:) - muuv*detrend(uvwplat(1,:));
    
    Cva = cov(detrend(uvw(2,:)),detrend(acc(2,:)));
    muva = Cva(1,2)/Cva(2,2);
    uvw(2,:) = uvw(2,:) - muva*detrend(acc(2,:));
    Cvv = cov(detrend(uvw(2,:)),detrend(uvwplat(2,:)));
    muvv = Cvv(1,2)/Cvv(2,2);
    uvw(2,:) = uvw(2,:) - muvv*detrend(uvwplat(2,:));
    
    Cwa = cov(detrend(uvw(3,:)),detrend(acc(3,:)));
    muwa = Cwa(1,2)/Cwa(2,2);
    uvw(3,:) = uvw(3,:) - muwa*detrend(acc(3,:));
    Cwv = cov(detrend(uvw(3,:)),detrend(uvwplat(3,:)));
    muwv = Cwv(1,2)/Cwv(2,2);
    uvw(3,:) = uvw(3,:) - muwv*detrend(uvwplat(3,:));
    
    % Now rotate into streamlines, etc...

Finally, it is important to note again that motion corrected winds in these files are
relative.  That is, they are in ship coordinates, corrected for high frequency motion
but not adjusted for ship heading or ship speed.

**************************************************************************************

The folder UH_Wind_RevB_decorrPlots contains png plot files of wind spectra and
cospectra for each hour.  These may be examined for QA/QC purposes.  In general,
the quality of the stream-wise wind stress cospectrum (u'w' cospectrum) is a good
indication of the effectiveness of motion corrections.  RevB plots for the
decorrelated wind show an improvement over RevA.  The utility of applying
the additional decorrelation step is still under discussion, and user comments are
welcomed.

**************************************************************************************

The file UH_Wind_RevB_decorrStats.txt contains hourly mean statistics from the
motion correction script.  Platform velocity variances are useful to filter out
hours with excessive motion.

u* is computed from rotated relative winds as follows (MATLAB).  Result is
unreliable when ship is underway and wind is not in-sector (+/-75 deg RWdir)

    %compute u*
    Cuw = cov(detrend(uvw(1,:)),detrend(uvw(3,:)));
    Cvw = cov(detrend(uvw(2,:)),detrend(uvw(3,:)));
    ustar = (Cuw(1,2)^2 + Cvw(1,2)^2)^(1/4);

The file contains the following variables:

	TimeStampStr: 	Date/Time UTC as string:  MM/DD/YY HH:MM:SS.sss
	Time_GPS_sec: 	Time in seconds from 0000 hrs, 1/1/2013
	uPlatVar: 		Variance in platform velocity for fore-aft axis
	vPlatVar: 		Variance in platform velocity for stbd-port axis
	wPlatVar: 		Variance in platform velocity for vertical axis
	ustar: 			u* in m/s
	RelWdirMean: 	Mean hourly relative wind direction, starboard positive.
	RelWdirVar: 	Variance in relative wind direction
	Azimuth: 		Azimuth angle from rotation into streamlines (== RelWdirMean)
	Tilt: 			Tilt angle from rotation into streamlines
	Umean_m_s: 		Mean wind velocity along streamline
	muua: 			Decorrelation coefficient (mu) for U acceleration
	muuv: 			Decorrelation coefficient (mu) for U velocity
	muva: 			Decorrelation coefficient (mu) for V acceleration
	muvv: 			Decorrelation coefficient (mu) for V velocity
	muwa: 			Decorrelation coefficient (mu) for W acceleration
	muwv: 			Decorrelation coefficient (mu) for W velocity
	
Note: the rotation and decorrelation steps are done on an hourly mean basis.
Therefore, they may be adversely affected if ship's heading or relative wind direction
changes greatly during the hour.  The best motion corrections are for hours with
minimal variance in heading and RelWdir.

**************************************************************************************