Re: Ambiguous coordinates on T axis

[Xiaobiao Xu <xbxu@xxxxxxxxxxxxxxx>]

well, I just tried the line you post, and I got different result!!!
xiaobiao
============================================================================
yes? set mode diag
yes? list/t="1-jan-1990":"31-dec-1990" a[l=@ave] - b[l=@ave]
 dealloc  dynamic grid (G002)          NORMAL    NORMAL    NORMAL    TWEEK
 getgrid EX#1     C:  5 I:    1    1  J:    1    1  K:    1    1  L:    1
1
 *** NOTE: Ambiguous coordinates on T axis: A[L=@AVE] - B[L=@AVE]
 allocate dynamic grid (G001)          NORMAL    NORMAL    NORMAL    TDAY
 strip limits reconciliation : EX#1
 eval    EX#1     C:  5 I: -111 -111  J: -111 -111  K: -111 -111  L:    1
365
 strip --> A[T=01-JAN-199000:00:31-DEC-199000:00@AVE,D=0]
 found   A        M:  3 I: -111 -111  J: -111 -111  K: -111 -111  L:    1
365
 doing --> A[T=01-JAN-199000:00:31-DEC-199000:00@AVE,D=0]
 strip --> B[T=01-JAN-199000:00:31-DEC-199000:00@AVE,D=0]
 found   B        M:  4 I: -111 -111  J: -111 -111  K: -111 -111  L:    1
314
 doing --> B[T=01-JAN-199000:00:31-DEC-199000:00@AVE,D=0]
             VARIABLE : A[L=@AVE] - B[L=@AVE]
             TIME     : 01-JAN-1990 00:00 to 31-DEC-1990 00:00
          8.514
==============================================================================

On Thu, 2 Oct 2003, Ansley Manke wrote:
> Hi again Steve,
> I take back the first part of my previous answer.  The calculation:
> 
> yes? shade/T="01-JAN-1994":"12-DEC-1994" \
>   aglt[D=1,L=@AVE]-aglt[D=2,L=@AVE]
> 
> does not in general compute the average from 01-JAN-1994 to
> 12-DEC-1994 of the two variables and subtract them.  You
> always need to specify the region for each variable separately
> in square brackets to get the calculation right, or regrid one dataset
> to the time axis of the other before doing the computation
> 
> yes? shade aglt[D=1,T="01-JAN-1994":"12-DEC-1994"@AVE]- \
>     aglt[D=2,T="01-JAN-1994":"12-DEC-1994"@AVE]
> 
> or
> yes? let algt2 = algt[d=2,gt=algt[d=1]]
> yes? shade/T="01-JAN-1994":"12-DEC-1994" \
>      aglt[D=1,L=@AVE]-aglt2[L=@AVE]
> 
> Try this: make up a couple of time series variables on two
> different time axes, and use MODE DIAGNOSTIC to see what
> Ferret makes of the calculation. Look at the lines that start with
> "doing -->" to see that the average for variable a is correct, but
> the averaging for variable b does not have the right time range.
> Ferret cannot make sense of the command.
> 
> yes? def axis/t="1-jan-1990":"31-dec-1990":1/units=days tday
> yes? def axis/t="1-jan-1989":"31-dec-1994":7/units=days tweek
> 
> yes? let a = sin(t[gt=tday]/10) * t[gt=tday]/100
> yes? let b = cos(t[gt=tweek]/20)* (1-t[gt=tweek]/100)
> 
> yes? plot a   ! Plot the variables to load them into memory
> yes? plot b
> 
> yes? set mode diag
> yes? list/t="1-jan-1990":"31-dec-1990" a[l=@ave] - b[l=@ave]
> 
>  dealloc  dynamic grid (G002)          NORMAL    NORMAL    NORMAL    TWEEK
>  getgrid EX#1     C:  5 I:    1    1  J:    1    1  K:    1    1  L:    1    1
>  *** NOTE: Ambiguous coordinates on T axis: A[L=@AVE] - B[L=@AVE]
>  allocate dynamic grid (G001)          NORMAL    NORMAL    NORMAL    TDAY
>  strip limits reconciliation : EX#1
>  eval    EX#1     C:  5 I: -111 -111  J: -111 -111  K: -111 -111  L:    1  365
>  strip --> A[T=01-JAN-199000:00:31-DEC-199000:00@AVE,D=0]
>  found   A        M:  3 I: -111 -111  J: -111 -111  K: -111 -111  L:    1  365
>  doing --> A[T=01-JAN-199000:00:31-DEC-199000:00@AVE,D=0]
>  strip --> B[T=01-JAN-198900:00:31-DEC-198900:00@AVE,D=0]
>  found   B        M:  4 I: -111 -111  J: -111 -111  K: -111 -111  L:    1  314
>  doing --> B[T=01-JAN-198900:00:31-DEC-198900:00@AVE,D=0]
>              VARIABLE : A[L=@AVE] - B[L=@AVE]
>              TIME     : 01-JAN-1990 00:00 to 31-DEC-1990 00:00
>          -0.1186
> 
> 
> 
> Ansley Manke wrote:
> 
> > Hi Steve,
> > When you do a calculation over time when there are two different
> > time axes involved in the expression, Ferret gives this informational
> > message just to let you know that the two time averages are different.
> > It's using each variable's time axis and is doing the calculation you
> > expected. You could make it absolutely clear what you're doing by
> > specifying the time interval as a qualifier for each variable, in square
> > brackets.  (I use the backslash continuation character here)
> >
> > yes? shade aglt[D=1,T="01-JAN-1994":"12-DEC-1994"@AVE]- \
> > aglt[D=2,T="01-JAN-1994":"12-DEC-1994"@AVE]
> >
> > See "region" in the Users Guide index; this is the third way to specify
> > a region and is almost always the best way to be sure you're doing
> > what you think you're doing. And, this is also the way to do it if you
> > want to compare averages over two different time intervals:
> >
> > yes? shade aglt[D=1,T="01-Jan-1994":"12-Jun-1994"@AVE]- \
> > aglt[D=2,T="13-Jun-1994":"12-Dec-1994"@AVE]
> >
> > Ansley Manke
> >
> > Steve Knox wrote:
> >
> > > How can I avoid this ambiguity on the time axis when I'm dealing with 2
> > > datasets?
> > > I guess it's confused since it dosen't know which T axis to average over.
> > > In this case the time periods are the same so the ambiguity doesn't
> > > matter, but if I want
> > > to specify  different time periods in each dataset how can I do this?
> > >
> > > yes? use agltHistoric.cdf
> > > yes? use agltCENT.cdf
> > > yes? shade/T="01-JAN-1994":"12-DEC-1994" aglt[D=1,L=@AVE]-aglt[D=2,L=@AVE]
> > >
> > > *** NOTE: Ambiguous coordinates on T axis: AGLT[D=1,L=@AVE]-AGLT[D=2,L=@AVE]
> > >
> > > Thanks for any help on this.
> > >
> > > Steve Knox
> > > NREL
> > > Colorado State University
> >
> > --
> > Ansley Manke  Pacific Marine Environmental Laboratory  Seattle WA
> 
> --
> Ansley Manke  Pacific Marine Environmental Laboratory  Seattle WA
> 
> 

 xiaobiao xu
 MPO/RSMAS UM, 4600 Rickenbacker Csw.
 Miami,FL,33149 
 Phone:(305)-361-4628(O); 595-7844(H)
 E-mail:xbxu@rsmas.miami.edu