meandrift_world.mat - Matlab file containing large-scale currents derived from
drifter observations.

For details of this calculation, see Lumpkin and Johnson (2013).  In summary,
the drifter observations are mapped to a model within bins that are oriented
and shaped according to the variance ellipse within the bin.  The model is:
observations = [1 , SOI, sin(t*2*pi), cos(t*2*pi), sin(t*4*pi), cos(t*4*pi) ...
       		x, x.^2, y, y.^2, x.*y] * coefficients + noise
where SOI is a five-month lowpass Southern Oscillation Index,
t is time in fractional year (e.g., 15 Jan is approx. 0.5/12),
x is the east-west distance (degrees) from the bin center,
y is the north-south distance (degrees) from the bin center,
"coefficients" are given in Ubar, Vbar, etc. (see below),
and "noise" is the residuals ... eddy speed u' in the case of zonal velocity.

This Matlab file contains the following variables:

Lat: vector of latitudes (degrees)
Lon: vector of longitudes (degrees)
NumObs: number of observations in bin (drifter days per square degree)
U, V, Uek, Vek, SST: monthly climatological zonal (U) and meridional (V) 
	velocity (m/s), zonal and meridional Ekman velocity (m/s),
	and sea surface temperature (degrees C).
	Each of these matrices has dimension length(Lat) x length(Lon)
	x month x 2.  Set the final index to 1 for the values, and
	to 2 for the associated standard error bar.
Ubar, Vbar, Uekbar, Vekbar, Tbar: 11 individual coefficients of model fit to 
	drifter observations.  Each of these matrices has dimension length(Lat) x 
	length(Lon) x 11 x 2.  Set the final index to 1 for the coefficient,
	and to 2 for the associated standard error bar. 
	Examples:
	Time-mean zonal speed at Lat=i, Lon=j: Ubar(i,j,1,1) +- Ubar(i,j,1,2).
	Climatological January SST at i,j:  
		t=0.5/12;
		model=[1+0*t, 0*t, sin(t*2*pi),cos(t*2*pi),sin(t*4*pi),cos(t*4*pi)];
		jan_temperature = model*squeeze(Tbar(i,j,1:6,1));
		error_jan_temperature = model*squeeze(Tbar(i,j,1:6,2));
Up2bar, Vp2bar, UpVpbar: <u'^2>, <v'^2> and <u'v'> (m^2/s^2), where these
	primed terms are residuals with respect to the model fit.