[GRASSLIST:1426] Centres of grid cells

[David Orme]

Hi,

Can someone suggest an answer to this mismatch between GRASS 5.0.2 and 
ArcMap 8.2?

A colleague and I are describing global distributions using a raster 
grid in a Behrmann projection (CEA with latitude of true scale at 30 
NS). For the purposes of converting geographic data into the grid we 
want to know the geographic coordinates of the centres and boundaries 
of the cells in the raster grid. More technical details below but as a 
summary:

I originally worked these out in Grass:
r.to.sites - to convert the raster data to a series of sites occurring 
in the centre of each cell.
s.proj - to drag the resulting sites file over into a geographic 
location from the behrmann location (both of which use the wgs84 
ellipsoid).
This gave me a list of lat/long values for the centres of the equal 
area grid.

My colleague then did the 'same' thing in ArcMap. I don't have the 
precise details but a shapefile containing square polygons (to form a 
grid) was converted to point centroids and then reprojected into 
geographic. Again both shapefiles use wgs84. The resulting list of 
lat/long values does not match the ones resulting from GRASS.

I would expect a mismatch if the 'grid' shapefile were reprojected and 
then the centroids of the transformed grids were found, but I wasn't 
expecting mismatch from the reprojection of point/sites.

Thanks,
David

Some details:
The behrmann grid has a resolution (NS and EW) of 96.486268 km (it 
matches to a degree at the latitude of true scale, before anyone asks!) 
and has a global coverage (360 cells EW by 154 cells NS - this goes 
above and below 90 degrees). The NS centres of the cells (returned by 
r.to.sites) in the Behrmann projection are therefore: ((the number of 
cells N or S * 96.486268) - 48.243134). To take a mismatch at the 
extremes as an example, Row 2 is centred on 7188.226966 km N - s.proj 
returns a site latitude of 82.48556 decimal degrees but arcmap reckons 
on a longitude of 82.85535 decimal degrees. All my estimates are 
correspondingly smaller than my colleagues.