G'day,
A convention for unused vertices would be useful.
The unstructured grids that I have tested handle
this in a sub-optimal way.
Most put the valid vertices first, and copy the
last valid vertice into all the unused vertices.
In other words, they represent unused vertices as
duplicates of the same vertice.
This approach requires tools (like ncremap/ncks)
to count the number of non-duplicated points
to determine the vertice number for each cell.
This is sub-optimal and error-prone.
I propose that the _FillValue attribute, if any,
of relevant CF bounds variables (e.g., lat_bounds,
lon_bounds) indicate unused vertices.
Although a _FillValue could theoretically appear
in any vertice, it would be most tool-friendly
if all the valid vertices were grouped together
at the beginning of the vertice array, rather
than having unused vertices interspersed throughout.
This would make it less cumbersome to read the
all useful vertices into an array because tools
could stop reading the vertices upon encountering
the first _FillValue. Are there good reasons
for unused vertices in the middle of an array?
If not, I further propose that valid vertices
be grouped together at beginning of the
vertice array.
Charlie
> Hi all,
>
> For unstructured grids when some cells are quadralaterals and others are
> pentagons, the convention stipulates that bounds should be dimensioned
> (n,5) so that the lat and lon locations of all 5 penatagon vertices can
> be recorded. My question is what to do about the extra vertex for the
> quadralaterals? Should we define a "missing_value" and store it in the
> unused vertex location?
>
> If so, perhaps we should make this clear in the convention document.
>
> best regards,
> Karl
--
Charlie Zender, Earth System Sci. & Computer Sci.
University of California, Irvine 949-891-2429 )'(
Received on Wed Apr 06 2016 - 12:32:30 BST
