focal/block statistics in combination with (polyline)barriers

[FreekBusschers]

Dear Members,

I’m looking for a way to apply focal/block statistics in combination with (polyline)barriers. I have a bottom raster of a geological unit that is tectonically segmented (see attachment). I want to select the highest raster values but the algorithm should not be allowed to look ‘beyond the fault line of each block’’, i.e. it should only use the values on the specific tectonic blocks for calculations. I observed that for example ''Contouring'' can be done using barriers, but is it also possible for focal/block statistics? And if not' an anyone create this for me?

I'm new in GeoNet so actually I don't know whether this is the correct group to post this.

Cheers, Freek Busschers, the Netherlands

[DanPatterson_Retired]

You might want to assign the breaks a nodata value, then 'buffer' (using Expand ) those zones by half the moving window size prior to running the focal statistics.  In theory the expanded breaks would be detected and assigned nodata and the moving window wouldn't pick up any values as it moves across the raster.  So for a 3x3 moving window, an Expand of 1 would results in a 3 cell wide break 

[FreekBusschers]

Dan, thanks for your reply! The only problem is that I have to do a very rough statistical smoothing (I should have indicated that in my text). My cell size is 100m, my moving window would be something like 3 by 3km (30 by 30 cells) meaning it would certainly look beyond the expanded break...... 

[DanPatterson_Retired]

In a situation like that then, you would have to do that post processing and obliterate cells 15 cells on either side.

What is the purpose for such a coarse smoothing?  Have you tried to see if there are differences in results with incremental kernels? ie 3x3, 5x5, 7x7....

There may be a threshold where you could get away with a smaller kernel.  This may sound counter-intuitive, but it would be worth a shot if the 'no data from the other side' is critical