Re: R: Re: [ferret_users] Spatial average on irregular domain

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi
You're right to look into that.  Using @SUM should work as you say. Or 
if the underlying coordinate axes (the x and y axes used by the grid of 
the 2D lon, lat, and sst variables)  are just index values, then the 
area calculation that Ferret does will treat each grid cell as being the 
same size.

Ansley

On 11/5/2011 10:05 AM, markus.mingel@xxxxxxxxx wrote:
> Dear Ansley,
> thanks for the suggestion, it works and I am able to create the mask. However
> I still have a question, since I am not sure about the averaging. What I did
> is:
>
> ! Create Mask
> let Mask_Tropics = if lat gt -30 and lat lt 30 then 1
>
> !  Mask the variable also multiplying it by grid area
> New_SST = Mask_Tropics * Area * SST
>
> ! Compute and Plot spatial mean
> plot  New_SST[i=@ave,j=@ave]
>
> I think last step is wrong since ferret attempts to compute a "fake" area of
> cells. So how can I avoid it accounts for the area? Maybe  could I use @sum
> instead of @ave and then divide by the number of points that fall inside the
> sub-region?! If so, how can I compute the number of points where Mask_Tropics =
> 1?
>
> Thanks for help
> Regards
> Markus
>
> > ----Messaggio originale----
> > Da: ansley.b.manke@xxxxxxxx
> > Data: 03/11/2011 17.40
> > A: "markus.mingel@xxxxxxxxx"<markus.mingel@xxxxxxxxx>
> > Cc:<oar.pmel.ferret_users@xxxxxxxx>
> > Ogg: Re: [ferret_users] Spatial average on irregular domain
> >
> > Hi-
> > This is not a simple question, is it.  The size of the grid cells on a
> > curvilinear grid may vary from cell to cell, so to compute an accurate
> > average you would need to find a way to compute the area of the cells.
> > On a rectilinear grid, Ferret computes the actual size of the grid cells
> > using a simple latitude correction, but it does not have a means of
> > doing that on a curvilinear grid.  Does the dataset have variables that
> > describe the area of cells? Some datasets do include that information.
> > If so then we could find a way to create a mask using the curvilinear
> > coordinate variables, along the lines of
> >
> > let mask = if curv_lat gt -30 and curv_lat lt 30 then 1
> >
> > and compute the average of the data multiplied by that mask. Another
> > option would be to regrid the data to a rectilinear grid and take the
> > average of the regridded field over the region.
> >
> > Ansley
> >
> > On 11/3/2011 6:09 AM, markus.mingel@xxxxxxxxx wrote:
> > > Dear All,
> > > I am trying to perform a spatial average on irregular domain.
> > >
> > > I know how to do that on regular domains:
> > >
> > > let regional_ SST = SST[x=-180:180@ave,y=-30:30@ave]
> > >
> > > however this command does not work or curvilinear grids. Could you please
> > > suggest me how to compute spatial averages in such case?
> > >
> > > Thanks for help
> > >
> > > Regards
> > > Markus