How to do 1D objective analysis

[Lev Tarasoff <lev@xxxxxxxxxxxxxxxxxxxxxxxxxx>]

Thanks to Ansley, Mark, and Steve, here's how to effectively do
1D scatter2grid on say a timeseries with uneven delta time (eg icecore
profile)

Assume mydat.dat has NDAT entries, and the row form is: time, data
then use define axis/from_data as follows
................
define axis/x=1:NDAT:1 x1
define grid/x=x1 grin
FILE/VAR='t1,var1'/GRID=grin mydat.dat
define axis/from_data/t/name=tg1 t1
define grid/t=tg1 gtg1
FILE/VAR='t2,var1'/GRID=gtg1 mydat.dat
................

you can then use standard regridding to put this time series onto a
regular time grid. One caveat, if you are going to a much coarser grid,
you should first regrid onto a fine regular grid with a smoothing
transformation otherwise the coarse regridding will just use adjacent
datapoints for the regridding which can be a problem if the timeseries has
lots of high frequency noise. Ie. suppose that the finest timestep for the
input data is dTin and for the desired grid is dT, then first regrid onto
a fine grid with this timestep (say gtgf) using for instance a Hanning
smoother using dT/dTin=NH for the box size
................
let var1f=var1[g=gtgf]
let var1fs=var1f[t=T0:Tf@SHN:NH]
let varfinal=var1fs[g=gfinalregulartimegrid]
................


cheers

Lev