I am working in ArcMap 10.3.1 with Spatial Analyst tools. I have a raster with some large contiguous areas with NoData. I want to assign the floating point value of the closest cell to each NoData cell. This would be similar (as I understand it) to using the Euclidean allocation tool, but the return value (i.e., output raster) for that tool is integer type, not the floating point type which I need. I want to do this only for areas with NoData, w/o affecting values in cells surrounding the NoData areas. The attached image shows a small part of the study area (colored areas have values, areas w/ DEM background have NoData). A similar question was asked here, but again, I want results to be floating point. Any ideas?
Sadly the Nibble tool requires an integer input. However, you could create an integer grid by multiplying you data by a factor equivalent to your acceptable scale. So, if you can accept, say five decimal places, create an integer grid from your raw input raster by multiplying by 100000.
Perform the nibble and then divide by the same factor. Obviously this will truncate your decimal places and you may experience some rounding errors, so the bigger the factor you can use the, the better - within the limits of your data and the pixel depth of your raster. Given that you could use a 32 bit or even 64 bit raster - I expect you have a lot of scope. Wrap the process in Model Builder and you have a one-click solution.