nx=2160; ny=nx*13; nz=90;

SSH/
	ssh_${YYYYMMDD}.bin	[nx ny]
				daily mean
		
	ssh_${YYYYMM}.bin	[nx ny 3]
				monthly mean +
				monthly mean square of hourly +
				monthly mean square of daily
	ssh_year.bin		[nx ny 3]
				yearly  mean + 
				yearly  mean square of hourly + 
				yearly  mean square of daily
THETA/				

	THETA/Theta_${YYYYMMDD}.bin	[nx ny nz]
				daily mean
	THETA/Theta_${YYYYMM}.bin	[nx ny nz 3]
				monthly mean +
				monthly mean square of hourly +
				monthly mean square of daily
	THETA/Theta_year.bin		[nx ny nz 3]
				yearly  mean +
				yearly  mean square of hourly +
				yearly  mean square of daily

SALT/U/V/W
	similar to THETA/

known issues: due to rounding error, when calculating standard deviation
		std_hr = ssh_mn_sq_hr - ssh_mn_mean.^2;
		std_dy = ssh_mn_sq_dy - ssh_mn_mean.^2;
		there are a small portions of negative std_hr/std_dy
		especailly for SALT (even after SALTanom treatment):
		ii=find(hFacC>0);
		std_hr(ii) = ssh_mn_sq_hr(ii) - (ssh_mn_mean(ii)-35).^2;
		std_dy(ii) = ssh_mn_sq_dy(ii) - (ssh_mn_mean(ii)-35).^2;
		


======
TIDES/
NX=35159453; NT=24*365=8760; nc=8;
	TIDE_SSH.bin (TIDE_SSH_NxT.bin) [NX NT]
			NT houlry SSH @ NX wet points
	TIDE_SSH_TxN.bin	[NT NX], ie, transpose of TIDE_SSH_NxT.bin

	TIDE_amp_pha_llc2160_sub2llc540.mat
		Amp540	    [nc nx/4 ny/4]
		Gph540      [nc nx/4 ny/4]
				from subsampling of TIDE_SSH_TxN.bin
				(for quick sanity check before full domain)
	TIDE_amp_pha_llc2160.mat
		Amp2160	    [nc nx ny]
		Gph2160     [nc nx ny]
				from full domain of TIDE_SSH_TxN.bin

	tpxo9_llc2160.mat has
	      	Amp2      [nc nx ny]     amplitude (m)
      		Gph2      [nc nx ny]	 phase (0-360 deg)
	tpxo9_llc2160_sub2llc540.mat has
		Amp3      [nc nx/4 ny/4]
		Gph3      [nc nx/4 ny/4]
				subsampling from Amp2/Gph2 (every 4x4)


		nc(=8) constituents are:
##components   = M2       S2     N2       K2       K1      O1       P1       Q1
##periods (hr) = 12.4206  12     12.6583  11.9672  23.9345 25.8193  24.0659  26.8684