I am using ArcMap 10.2.2, and I need to aggregate categorical data from 30m up to 90m and 900m resolution for species distribution modeling. The algorithms available in resample (nearest neighbor, bilear, etc.) are not appropriate for my purposes, and the Aggregate tool does not have a "majority" option. Thus, I think I have to use the Block Statistics tool. However, I am getting a strange, recurrent error where some cells in the output raster show up as noData. Has anyone else seen this problem? I got the same effects when working with both fgdb rasters and geotiffs. The same apparently random cells show up as noData over multiple runs. The input data does not have any noData cells in the areas where I'm seeing the problem.
As a work around, I can do this in R using the aggregate function in the raster package. Their algorithm is inexplicably slow, though, and it's more convenient for me to work in Arc if possible. Here are the messages from the results panel, just in case they're helpful.
Executing: BlockStatistics NLCD_2011_WV_WGS_1984.tif C:\Users\mosteele\Desktop\aggregation_test\NLCD_2011_WV_90m_ArcMap_version2.tif "Rectangle 3 3 CELL" MAJORITY DATA
Start Time: Thu Jun 25 11:45:37 2015
Succeeded at Thu Jun 25 11:45:40 2015 (Elapsed Time: 2.46 seconds)
Output (noData in red):
Upon further inspection, I realized that the nodata cells only occurred where there was a tie for majority. I contacted an Esri support person who told me that this is the expected behavior.
I can understand why the algorithm was created like this, but I'd argue that it would be much more useful to have two options - one that would set ties to nodata and one that would, say, break ties randomly. I needed the output surface for species distribution modeling and a version with a bunch of nodata is useless to me. I'm also unsatisfied with Esri's other methods of resampling categorical data, such as the majority option in the Resample tool or resizing by snapping, as they drop a lot of information from the original raster by looking only at a limited window around the centroid of the new cell.
Fortunately, I was able to create the surfaces I needed using R - specifically, the aggregate tool in the raster library.