Convert coordinates

262
6
02-10-2014 03:22 AM
PV
by
New Contributor III
Hi,
I have a problem with converting the WFDEI data set i use (meta information can be found here: http://www.eu-watch.org/gfx_content/documents/README-WFDEI.pdf[/url]) to the correct projected coordinate system in meter format (I need it in USA_Contiguous_Albers_Equal_Area_Conic_USGS; geographic coordinate system GCS_North_American_1983)

The WFDEi data comes in lat long (degrees) in a .nc format but i am not able to convert it to my desired projected coordinate system, since the projection is listed as unknown in arcgis and since i could not find any information about it in the meta data.

I got these informations from the .nc file but I dont really know how to interprete them:

Generated grid information:
---------------
gridtype = lonlat
gridsize = 6477
xname = lon
xlongname = Longitude
xunits = degrees_east
yname = lat
ylongname = Latitude
yunits = degrees_north
xsize = 127
ysize = 51
xfirst = -125.37598
xinc = 0.5
yfirst = 24.562113
yinc = 0.5
---------------

Can anyone help me with that?
Thanks alot!!
0 Kudos
6 Replies
markdenil
Regular Contributor II
the readMe says:
All WFDEI data files include grid-box centre longitude and latitude (on a regular longitude-latitude 720 x 360 grid).

"a regular longitude-latitude .. grid" is likely a geographic lat/long grid; on what datum... who knows? try a sphere.

the "Generated grid information"
reinforces that impression: gridtype = lonlat
The 'gridsize = 6477' squares with the xsize = 127 and ysize = 51
xinc = 0.5 and yinc = 0.5, so the cells are half degree

The center of the lower left cell is xfirst = -125.37598, yfirst = 24.562113
(remember the "..data files include grid-box CENTRE longitude and latitude.." from the readme)

so, subtract 2.5 from the given x and y to find the lower left corner.

That is all pretty simple; the datum may be a problem
(self-identified scientists types tend to ignore such things)
give it a shot defining the datum as wgs84
or GCS_North_American_1983
and see how it lines up once you have it projected to your target.

If it is off, try defining the grid on a sphere (size is, again, anyone's guess)
if you do that, you may have to create a custom (dummy) transformation with all the values at 0
to get from the sphere to the NA83 spheroid.

if it is still off a consistant distance, you can build in a coordinate shift.

This may sound like it can be a little snakey (a lot snakey), but people that build these types of files often don't pay much attention to mapping conventions; they just want things to be consistant within their own user community....
0 Kudos
PV
by
New Contributor III
Thank you very much for your helpful answer!!!
I think the coordinate shift is the only solution I have left over, however I couldnt make it work yet.
0 Kudos
markdenil
Regular Contributor II
There is a little work-around routine to do that.

project the data to the target projection
The one where the features are mis-aligned
(each feature is, say 2km too far south and 5 km too far west)

replace the prj on the data (define the projection) with a proj using a
false easting and northing that makes the data line up.

Now re-project the data back into the original target projection
(the one without your adjusted easting and northing).

This should shift everything to where you want.

Elegant? no, not elegant in the least....
0 Kudos
PV
by
New Contributor III
I managed to get the data close to my desired projection but however what kind of false easting I apply, I cannot align it.
I attached you a screenshot on how it looks like now.. can u give me a guess on what number I have to enter?
Thank you a lot!
0 Kudos
MelitaKennedy
Esri Notable Contributor
Very rough guess/estimate based on the image.

false easting: -36000 meters
false northing: -71000 meters
0 Kudos
PV
by
New Contributor III
It seems to work!!! Thank you a lot!
(Althoug I had to take the positive values for northing and easting, without "-")
But how can I be 100% sure that the data is aligned perfectly?

Also I am wondering if the raster still can be used for analysis and overlays with other raster since the squares
are not actual squares anymore now because of the whole reprojection process?

I am sorry for all these questions, but I am not an expert in ArcGIS!
0 Kudos