Re: [ferret_users] FILL w/o interpolation at coast line?

[Steve Cousins <cousins@xxxxxxxxxxxxxx>]

On Fri, 27 Jun 2008, martin.schmidt@xxxxxxxxxxxxxxxxx wrote:

> Hi Steve,
>
> guess your problem is, that you have zero and not missing values in your
> land points? Then you may want to replace such data with missing values
> before you
> make a plot. Possibly this helps.

Hi Martin,

The land values are not zeroes. The metadata for this variable is:

	short temp(time, zpos, ypos, xpos) ;
		temp:valid_range = -5.f, 30.f ;
		temp:long_name = "Temperature" ;
		temp:units = "Celsius" ;
		temp:scale_factor = 0.0005340739f ;
		temp:add_offset = 12.5f ;
		temp:missing_value = -99999.f ;

After the scale factor, offset and short integer modulo arithmetic the 
missing value in the file becomes -25184 which translates back to 
-0.950117.


> Another hint - there is a function fill_xy, which extrapolates data as
> smooth as possible. It fills missing values with average values from
> valid neighbour points.
> It is of some help, if you want to interpolate and extrapolate data and
> have some of your interpolation grid points near land.
>
> use coads_climatology
> fill/l=10 sst[d=1]
> SET MEMORY/SIZE=100
> go fland 5
>
> You will see some white areas near land.
>
> Now define a field of ones, needed for fill_xy
>
> let mask=0*missing(sst[l=1],1)+1
> let sst_fill=fill_xy(sst,mask,2)
>
> fill_xy fills missings, if the mask is 1 and does nothing, if mask is 0.
> "2" is the number of
> nearest neighbour filling cycles.
>
> Enhancing this value defines SST also around Antarctica. So be carefull.
>
> set win 2
> fill/l=10 sst_fill[d=1]
> go fland 5
>
> If you have zeros at land, you may proceed in two steps. First replace
> zeros by missings. Then you
> may extrapolate as described here.
>
> Please note, a nearest neighbour filler does about the same here. But
> again, first you should
> replace zeros (valid data) with missing values.
>
> Does this solve your problem?


I see no difference between the FILL TEMP and the FILL SST_FILL:

	http://rocky.umeoce.maine.edu/cousins/ferret

However, what is the purpose of fland here? I tried it but it complained 
about not having etopo5:

yes? go fland 5
 **TMAP ERR: non-existent or not on line
             Searched $FER_DESCR and $FER_DATA for etopo5(.des,.cdf,.nc)
set data etopo5
Command file, command group, or REPEAT execution aborted

Was this part of fill_xy technique you were describing or just an example.

Thanks for your help.

Steve