Re: [ferret_users] Re: [ferret_users] fill_between accuracy

[Steve Hankin <steven.c.hankin@xxxxxxxx>]

Hi Patrick,

They say, "You can't argue with success."  But I am going to, anyway. 
;-) Offering "2 cents" on this topic by way of discussion:

The process of interpolating in a plane is inherently ambiguous. I don't 
think (agree?) that one can **in general** guarantee precise alignment 
of plot lines and contour paths.  An alternative way that does ensure 
perfect accuracy is to use the segments of those lines, themselves, to 
define color-filled polygons.  There are Ferret scripts that take this 
approach.  Use of those scripts seems like the most reliable approach.

    - Steve

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

On 3/15/2017 12:58 PM, Patrick Brockmann wrote:
>   
> Le Mercredi 15 Mars 2017 17.28 CET, "Patrick Brockmann" <patrick.brockmann@xxxxxxxxxxxx> a écrit:
>   
> > Hi ferreters,
> >
> > I am wondering if the accuracy of the filling could be improved
> > in the following script.
> > I should perhaps regrid to a finer time step but I am afraid to
> > complicate the script unnecessary.
> I answer to my own question...
> With 2 additional lines, the problem is solved. No other way.
>
> My explanation.
> You need to define a finer axis time (days) and regrid the variable
> to it because the fills will be done by parts.
> Smaller they are, better the accuracy will be.
>
> So
>
> let a0 = uwnd[i=@ave,j=@ave]-uwnd_clim[i=@ave,j=@ave,T="01-JAN-1982":"31-DEC-1992"]
>
> to be replaced by
>
> let a0 = uwnd[i=@ave,j=@ave]-uwnd_clim[i=@ave,j=@ave,T="01-JAN-1982":"31-DEC-1992"]
> def axis/t="01-JAN-1982":"31-DEC-1992":1/units=days timeDays
> let a = a0[gt=timeDays]
>
> Happy ferreting.
> Patrick
>
>