Stacking overlapping rasters, keeping cell values on top?

Question

I have a land cover classification that consists of three files. One file is the LCC for the western half of my study area, another covers the eastern half (these two overlap in the middle), and the third represents a singe vegetation type which is present in both of the other two images (overlaps both). I would like to stack them western, eastern, single veg type, from bottom to top respectively, and end up with a raster that keeps only the top most cell values. Basically, I want a single image with the same appearance as the three when stacked. I'm working with ArcMap 10.2 and Erdas Imagine 2013 - so solutions in either software would be helpful.

Answer 1

There are several methods to combine rasters in ArcGis; generally ERDAS is faster at raster functions than ArcGis but I'm so out-of-touch with ERDAS that I can't even remember how to mosaic.

The first thing to consider is order of precedence. Lets assume that west is more important than east. Using the Mosaic to New Raster tool you can combine the three rasters into one. The only concerns are to get the rasters in the right order and use the appropriate overlap method, assuming you are using this as a tool and not in a python script either:

Use overlap first and select the vegetation type raster, west then east.

Use overlap last and select east, west then vegetation type raster.

NoData never overwrites data... so to ensure that top rasters don't mask lower rasters you must set the NoData value using the SetNull tool; for example if your raster contains classified data but 0 for nothing use SetNull to make the 0's into NoData.

Another option is to create a mosaic dataset, which doesn't combine the rasters but rather draws them in turn.. you can adjust the drawing order after creation until you're happy with the result then use Copy Raster to make it into a single raster dataset, this tool can also be launched by right clicking on the mosaic dataset and selecting raster to different format (or something like that).