Projection problem with 'ASCII to Raster' output

[steveSuperczynski]

Hello all,

I have model data that was converted to a raster from an ASCII file. This dataset covers essentially the eastern 1/3 of the CONUS,  Because this data has a set spatial resolution (12km) I used UTM coordinates and meters for units in the ASCII header.  The problem comes when attempting to project this data. When comparing the model output and a US shapefile, the data does not appear to be projected properly as the coastlines do not match up.  The reference point used to convert the text file is in UTM, however the coordinate system is still unknown.

I am no expert in GIS or map projections by any means, however my thought was that I should define the projection in the UTM zone that the lower-left corner is located (since my starting points were in UTM and my cellsize is in meters).  Does this seem reasonable or is there another step that needs to happen to project the data properly?

Thanks,
Steve

[steveSuperczynski]

sorry, forgot to upload a sample .txt file that I have been using. Here it is in a zip file.

[JeffreySwain]

If your original data was in UTM and you projected the data to a new coordinate system, then did you change datums?  If so did you specify a transformation in the map document?  You may need a transformation specified for the data to line up correctly. Please refer to this blog which discusses how to determine the proper transformation.

[MelitaKennedy]

Where did you get the data? Is there any information there about the coordinate system or the data's lower left coordinates in any coordinate system? It looks closer to USA Contiguous Albers Equal Area Conic USGS (projected coordinate systems\continental\North America) but you'll need to figure out the coordinates for the lower left corner in that coordinate system.

To see what I mean, add the converted raster and some reference data to ArcMap and set the data frame to USA Contiguous Albers Equal Area Conic USGS but modify (in 10.0, right click the name) it. Change these parameter values:

false easting: 210000.0
false northing: 2600000.0

You can see that the coastline matches better than if you set the data frame to NAD 1983 UTM 15N (the closest match to some reference data).

UTM shouldn't be used for this large an area.

Melita

[steveSuperczynski]

Well this was my first attempt at this. Unfortunately I am not sure what the coordinate system of the data provided to me is/was. It is gridded model output with a constant cellsize of 12km.  I do have a dataset with lat/lon values for each pixel which was also provided to me, but I am not sure exactly how to use that to map the raster data.

[MelitaKennedy]

If you can post the lower left corner all corner coordinates, I can convert them to the Albers definition and look at the results. I would say try to convert it directly to lat/lon, but deciding on what the equivalent cell size would be in degrees is problematical!

No possibility of asking the data provider for more information?

Melita

[steveSuperczynski]

The person I received the data from is no longer affiliated with our group, I have attempted to get in touch with him. I thought about attempting to convert everything to lat/lon but I came to the same conclusion.

Attached is the lat/lon data file I have. This is where I got the LLcorner coordinates from.

-Steve

[MelitaKennedy]

It's...reasonable. I don't know the data well enough to make a final judgment! Try changing these two lines:

XLLCORNER 11421
YLLCORNER 275000

and then setting the data frame to USA Contiguous Albers Equal Area Conic USGS to check it out.

Melita

[steveSuperczynski]

That does look better, though its still not quite there.  The problem here is its all a bit of guesswork due to the data itself and the resolution. The placement of the coastline is just a 'best estimate'. Which leads me to wonder...

Would it be possible to use the lat/lon dataset to somehow georeference the image?  Its either that or create control points and warp the image to match up properly...although I assume you can't use control points from one raster to warp another correct?

-Steve

[MelitaKennedy]

Based on the 'look' of the raster, I think it was built in Albers, or possibly Lambert conformal conic, but we just don't know the parameters. You can try to georeference directly to geographic (pseudo-Plate Carree), but it's going to be worse. Georeferencing via world/header file is in a sense just overlaying a set of xy lines over the existing raster. If you look at how the longitude and latitude lines should overlay it, I think you'll see that they're angled or curved. Now you could import it with a start point of lat/lon and then use the georeferencing toolbar, but that's going to warp the data to fit the reference data that you use.

Melita