Re: Ferret question

[Steve Hankin <hankin@xxxxxxxxxxxxx>]

Hi Mark,

Scanning your command lines I do not see where any interpolations are occurring. The command

FILE/var="buoy,yr,mo,da,hr,mi,airt,spl,uwind,vwind,sst1"/columns=11/grid=timegrid
"/disk46/mpickett/46030hr"

reads all of your variables in as irregularly-spaced time series. When you plot the data, the
default style is to connect the points, but there is no numerical interpolation occurring. (Try
PLOT/SYMBOL to mark the data points without a connecting line.)

If you would like to place this data onto a regularly-spaced time axis with gaps at the points
where no measurements exist there are several tools to help with this. If all of your points,
(where they do exist) lie exactly on a regularized sampling scheme, you can define the regular
axis using a standard DEFINE AXIS and regrid to this axis using GT=new_tax@XACT. This will
retain the data where it exists, only (missing values at points where input values are absent.)

If your points do not lie exactly on a regularized time sampling scheme you can still define a
new axis of arbitrary regular time spacing. Regrid to it using either GT=new_tax (linear
interpolation) or GT=new_tax@AVE (averaging). Then use the transformation @CDA and @CDB on your
original irregular time axis to define a mask (LET my_mask = ...) that will remove any points
where a valid measurement does not exist within limits that you specify in the mask definition.

I confess that I do not grok the lines in which you attempt to regrid the time series data to a
spatial grid ....

    - steve

====================================

mpickett@upwell.pfeg.noaa.gov wrote:

> Help!
> I'm a new user and my Ferret commands are very basic.  I'm comparing global
> wind models with buoy data and am having problems handling the missing data in
> the ASCII buoy files.  Ferret seems to want to interpolate across any gaps in
> time in the buoy data.  I would like to have missing data in the gaps so I
> don't have any "fake" data in my comparisons.
> My backup plan is to fill the gaps with missing data values using a Fortran
> program then bring it into Ferret. I am wondering if there is something simple
> I can do in Ferret to put in missing data.
>
> Here is the code I'm using.  The derivative method I'm trying to use only works
> to remove large gaps.
>
> define axis/t=1:35040:1 tax1
> define grid/t=tax1 readgrid
> set
> data/ez/grid=readgrid/skip=0/var="buoy,yr,mo,da,hr,mi,airt,spl,uwind,vwind,sst1"/columns=11
> "/disk46/mpickett/46030hr"
> show data
> let time=days1900(yr,mo,da)*24+hr
> define axis/from_data/t/units=hours/name=tax2/t0="01-jan-1900" time
> define grid/t=tax2 timegrid
> set
> data/ez/skip=0/var="buoy,yr,mo,da,hr,mi,airt,spl,uwind,vwind,sst1"/columns=11/grid=timegrid
> "/disk46/mpickett/46030hr"
> show data
> define axis/x=121.9w:121.89w:.01/units=degrees xaxis
> define axis/y=35.7n:35.71n:.01/units=degrees yaxis
> define axis/t="01-jan-1997":"31-dec-2000:18:00":01/units=hours tax3
> define grid/x=xaxis/y=yaxis/t=tax3 spacegrid
> let zero=0*(i+j)
> let u=uwind[g=spacegrid]/-100.+zero
> let v=vwind[g=spacegrid]/-100.+zero
> let sst=sst1[g=spacegrid]/-100.+zero
>
> !Use to remove any interpolations done by Ferret
> let du=u[t=@ddf]
> let uclean= if abs(du) lt 5e-6 then (-99.99) else u
> let vclean= if abs(du) lt 5e-6 then (-99.99) else v
> let sstclean=if abs(du) lt 5e-6 then (-99.99) else sst
> set var/bad=-99.99 uclean
> set var/bad=-99.99 vclean
> set var/bad=-99.99 sstclean
>
> save/format=cdf/file="/disk46/mpickett/buoy_46030.nc" uclean,vclean,sstclean
>
> Here is a sample of a ASCII buoy data file with a large time gap (some are only
> a few days and harder to detect).
>
> 46028 1997  7 17  0 50  144 10143  -865   541 1410
> 46028 1997  7 17  1 50  145 10141  -827   694 1410
> 46028 1997  7 17  2 50  145 10141  -636   785 1410
> 46028 1997  7 17  3 50  142 10139  -825   743 1410
> 46028 1997  7 17  4 50  141 10137  -750   750 1410
> 46028 1997  7 17  5 50  142 10135  -785   813 1410
> 46028 1997  7 17  6 50  141 10134  -797   743 1410
> 46028 1997  7 17  7 50  141 10129  -832   749 1410
> 46028 1997  7 17  8 50  141 10129  -848   530 1410
> 46028 1997  7 17  9 50  142 10131 -1008   328 1410
> 46028 1998  8 14 17 50  142 10164  -183   344 1600
> 46028 1998  8 14 18 50  144 10158  -290   196 1640
> 46028 1998  8 14 19 50  145 10156  -334   135 1640
> 46028 1998  8 14 20 50  145 10150  -547   221 1640
> 46028 1998  8 14 21 50  149 10147  -506   341 1640
> 46028 1998  8 14 22 50  151 10136  -357   457 1640
> 46028 1998  8 14 23 50  152 10135  -499   276 1640
>
> Thanks for any help anyone can provide.
> -Mark Pickett-
>
> Steve Hankin wrote:
>
> > Hi Mark,
> >
> > I'm afraid that you havn't provided enough details to know for sure what the
> > problem is.
> >
> > Can you send you question again, but include a fragment of the ASCII data
> > file, which shows how missing values and times are represented, and also
> > include your Ferret commands.
> >
> > May I suggest that you resubmit the question to ferret_users -- usually
> > someone will have a quick answer.
> >
> >     - steve
> >
> > ===========================
> >
> > mpickett@upwell.pfeg.noaa.gov wrote:
> >
> > > I'm stuck on a Ferret problem and was hoping you could give me some
> > > help.
> > > I'm loading ASCII buoy files to plot buoy winds over FNMOC model winds
> > > for comparisons.
> > > The buoy data has gaps when the instruments are broken.
> > > When I put the buoy data into a time grid, Ferret wants to interpolate
> > > across the missing times.
> > > I've tried to cancel mode interpolate and bad=-9999 but it still wants
> > > to fill those voids with a number.
> > > I'd like to fill the voids with "...." missing data.
> > > Do you have any ideas?
> > >
> > > My current plan is to use a Fortran routine on the buoy data to find the
> > > gaps and put in missing data.
> > > Then open it up into Ferret.
> > > Thanks,
> > > -Mark Pickett-

--
Steve Hankin
NOAA/PMEL, 7600 Sand Point Way NE, Seattle, WA 98115-0070
ph. (206) 526-6080 -- FAX (206) 526-6744