COMBINE Rasters different projection and resolution

[JulieSmith]

I am combining 3 raster datasets with the following information ...

Raster 1: Continuous Signed Integer, 16 Bit, Cell Size (463.3127, 463.3127), Linear Unit = Meter (1.0), User_Defined_Sinusoidal, 1 Band
Raster 2: Continuous Unsigned Integer, 8 Bit, Cell Size (0.0083, 0.0083), GCS_WGS_1984, 1 Band
Raster 3: Continuous Signed Integer, 16 Bit, Cell Size (0.00833333, 0.0083333), GCS_WGS_1984, 1 Band

I tried changing the projection of RASTER 1, but then it completely messed things up (I'm guessing because the cell size is so big - how do I change the cell size then?) Also, how can I change the 'bits'.

I'm getting very frustrated with us .... Any insight anyone may have would be great!

Thank you!

[JulieSmith]

So I was able to reproject (reprojected them to World_Sinusoidal), adjusted the cell size, and changed the linear measurement (to meter). But now when combining the 3 datasets the the attribute table comes up without any values. And actually when I reprojected the maps, the two that were projected to the sinusoidal disappeared (one raster was already on a sinusoidal projection.

Disappearing spatial data and missing attribute table ... any thoughts on why? This is driving me crazy!

[MichaelStead]

I know nothing about sinusoidal projection, but suspect your first problem is that "User defined sinusoidal" sounds suspicously like something modified from standard projection. If things are "disappearing" this would be my first guess as to why. Bring in some data of known projection, set your dataframe projection to the same as your known layers, then add your projected rasters and see what layers fall as they should. I suspect your two layers in GCS_WGS_1984 should reproject to the correct place on the globe and your sinusoidal home job will be off in space somewhere.

You can do processing that modifies cell sizes, but you can't "change" them any more than you can give an unprojected data set linear units. I have to assume you are omitting information regarding intermediate processes. Yes?

[JulieSmith]

Hi Michael,

Thanks for your reply. So I did as you suggested. Opened a new map, set the projection and added the rasters (reprojected rasters) so that they're all the same.

You're right, it was probably a projection issue because now they all show up with the same projection, same linear measurement, same datum, perfect! The only issue I'm having now is that when I 'combine' them using spatial analyst, it fails. The files are quite large (it says uncompressed is 33GB) because it's a global dataset so I wonder if it's too large of a dataset to combine? Is there a limit? If so, do you know if I can just divide the world into different regions and then combine. Then can I use mosaic to raster to merge all the datasets?

Thanks again!

[JulieSmith]

Michael, another question I had ...

I am changing the resolutions. I'm going from 30 arc second (~1km) to linear (500m) resolution. When I projected and 'resized' I chose 'majority,' but I wasn't sure how much of a loss/change of information that would be. Do you know if there's a way to measure how much information was lost/changed/adjusted for the cell size and projection change?

[MichaelStead]

There probably is a way to measure this. I know there is a whole set of tools in the toolbox for comparing datasets/tables/etc but I have not use them. It is going to vary on a global dataset. At the equator a degree is ~111 km and ~17 km at 80 degrees (while at the poles it obviously shrinks to 0, but this creates additional problems). 30 arc seconds IS almost a kilometer at the equator, but it is 140 m at 80 degrees.

If you just want something for display and are using v10 there are a whole pile of tools I just bascame aware of in the image analysis toolbar. You should be able to make a temporary mosiac in your mxd that may allow you to get around the file size issue. I think you can only have a 4G file in a XP machine with FAT32/NTFS file structure. I think GDB gets around this somehow (somewhat like a folder??). Anyways, you might see if you can't get a mosiac this route.

For a permanent mosaic you don't want combine, you want mosiac to new raster from data management>raster

[JulieSmith]

Hi Michael,

Thanks so much for your reply.

I just finished reprojecting the rasters (the final raster being 300GB!). And yes, I am using ArcGIS 10, so I'll try playing around with that as well as the mosaic tool as you suggested and I'll be crossing my fingers. I'll post my solution to here (hopefully I'll find a solution soon!)

Thanks again!

Cheers.

[MichaelStead]

300 G? That is some file. I saw something about bigtiff as a tiff format developed for massive imagery files, but that is crazy big. Is it in a GDB?

Oh and something else I forgot to mention.... that whole 111km at the equator to 17km at 80 degrees is only in the x plane. The variance is fairly small in the y plane going from the equator to the poles. Just look at lat long lines on a globe as a visualization aide.

[JulieSmith]

Michael, Thanks again, you've been a great help!

Yea, it's pretty big. I tested out combining it by clipping smaller regions (ie. USA) and it worked! Still takes a while, but it's working! I'm very excited! But, now I'm having a different issue, which I think has an easy solution, but having a hard time figuring out ...

When combining rasters, it automatically only combines the 'value' field of the raster field. I have a separate column (carbon), not the default 'value' field. The 'value' field in that raster does not give the information I need - so that's why I created a raster to get the 'carbon' field. I was wondering if there was a way to change either the field of combine setting - seeing that the combine tool only seems to combine the 'value' fields and no other.

I'm assuming there are two options.
1. Change the 'value' field to have the carbon values instead
2. Adjust the combine tool settings so it can combine the 'carbon field,' instead of the 'value'

Not sure how to do either of those though. Made some attempts, but it's getting late and been working on this for a while and so I may not be thinking as clearly as I could be. Will try again after some shut eye.

In the meantime, if anyone has some insight on this it's greatly appreciated. Thank you!

[JulieSmith]

a great tool: spatial analyst > reclass > lookup
http://help.arcgis.com/en/arcgisdesktop/10.0/help/index.html#//009z000000sn000000.htm

just completed the map - looks great!