The general scenario
A raster of landuse landcover data with 123 categories (unique pixel values), classifies habitat and ecosystem types in a statewide dataset at 10-meter resolution. For internal use, the 123 original values are assigned to a fewer number of more general classes. For most of the original values, it is as simple as using Lookup Raster to assign pixel values to the more general class. However, a subset of the original values need to be reclassified to the nearest neighboring land cover of a specific type before generalizing.
For example; the simple Lookups
| Original pixel value | Original definition | General category, lookup value |
| 30 | Upland Pond Depression Herbaceous Vegetation | Herbaceous Wetland |
| 41 | Miss. Alluvial Valley Wet Herbaceous Vegetation | Herbaceous Wetland |
| 116 | Ozark Riparian Wet Herbaceous Vegetation | Herbaceous Wetland |
| 117 | Interior Floodplain Wet Herbaceous Vegetation | Herbaceous Wetland |
The table above shows a sample of values that one-to-one are equivalent with a specific general category.
The question
What tool or workflow is appropriate for the following:
For cells with values a, b, c ; identify the nearest cell that has a cell value of d, e, or f; reclassify to that nearest neighbor.
Here, a, b, and c are original values that are very specific types of forested area that don't immediately fall within one of the more-general categories, so they should be "absorbed" by the nearest neighboring forest type, which do have a corresponding general category.
Example
Four pixel values (2, 14, 94, and 109) represent four types of Wooded Vegetation and Shrubland areas (the a b c mentioned above). Of the original pixel values, I have a long list of all the various Forest/Woodland/Shrubland types (the d e or f mentioned above).
A logical statement for this might say: For each cell where the value is 2, 14, 94, or 109 -- identify and reassign the output value to the nearest cell value found in the list of all the Forest/Woodland/Shrubland types.
---
I am thinking that the "Nibble" spatial analyst tool may be the direction; however, I do not have experience with this tool to fully understand how I might configure the Mask raster; so that all other values in the raster are unaffected -- only those four cell values should be reassigned to the nearest cell value (but where that nearest cell value could be one of any of the many types of wooded/forest/shrubland).
---
Thank you for your time and please allow me to clarify if you have questions. I am looking for someone more experienced with raster and spatial analysis that can clarify if the Nibble tool is correct or an alternative tool or workflow.
If Nibble is correct for this use-case how should I define the settings from the geoprocessing tool and set up the Mask raster?
