# Raster smoothing using ArcMap

I need to simulate a debris avalanche, but in my elevation raster the avalanche deposit is included so I'd need to take it off to simulate on a pre-avalanche-like surface.

Unfortunately the only data I have is the average (and supposed uniform) thickness of the deposit of 24m. I removed this thickness from my raster using raster calculator ("my raster" - "raster of tickness created from polygon to raster").

What I obtained is a raster without that thickness, but like a valley with sub-vertical walls! I need to smooth these edges to make it look more realistic (now it looks like a print with the shape of the deposit).

What I was able to find is the Filter tool which smooth these edges too mildly or the Smooth Line tool which could work on the contour lines.

Is there a tool to smooth directly the raster?

• possible duplicate of What raster smoothing/generalization tools are available?
– Aaron
Commented Jun 14, 2013 at 13:59
• @Aaron That's a good call; those methods can be applied to answer this question. But IMHO there may be better answers available which address some special aspects here. The present problem might better be characterized as "fairing" a raster into another rather than just "smoothing" or "generalization," although it does share aspects of the latter two. A good answer would use knowledge of the shapes of debris deposits. It would also increase the thickness in the middle to compensate for the thinning at the edges, maintaining the average at 24 meters. Commented Jun 14, 2013 at 14:43
• Any chance you can post an image of your raster? Commented Jun 14, 2013 at 17:11
• @Radar This is kind of the reverse of How to build an artificial dam?, which has some illustrations: given the result, which is a raster showing a dam, how could one reproduce the original DEM that doesn't have the dam? The dam (which is just an artificial "deposit" of geological materials anyway) plays the role of the avalanche deposit. This analogy helps make it clear why some knowledge of the likely 3D shape of the avalanche deposit (or the terrain beneath it) is critical to a good solution. Commented Jun 14, 2013 at 17:42
• Agreed - if you want to remove a 3D object from 3D space you need to know the dimensions of the object. An example image would help me wrap my head around the geometry of the object to be removed (although I can make an educated guess as to what slide deposit would look like - I'm a bit concerned it may not be as uniform as the original post states). Commented Jun 14, 2013 at 18:19

I would do three things:

1. draw a polygon around your depression so that it overlaps your walls by a distance of at least 24 m. You can play with the buffer distance after you see your first results if the results don't quite suit your fancy.

2. I would run a focal statistics tool with the above polygon set as a mask for processing extent. Probably Mean value with a 5x5 height/width cell value. This will smooth out just the area of the depression. (Note: you could use any number of smoothing raster tools here. It's just my preferred tool, and I think it will work)

3. I would run a boolean raster calculator to basically paste the value of your focal statistics calculation over your original raster. You should be left with a smoothed depression that only affects your area in question while leaving the rest of the raster untouched.

You could also make a ring-shaped polygon that only covers the vertical wall, and that way you'll preserve your values in the center of the depression as well.

The deposit thickness is unlikely to be of a uniform thickness. I would suggest downloading a DEM from a date before the avalanche and using it as a starting point.

If you cannot find a DEM for your area, it is possible to create such DEMs using SAR satellite imagery such as Sentinel-1 in the SNAP software, which is free. the achievable resolution is around 10 m/pixel

I think you need the Raster Generalization option: http://www.esri.com/software/arcgis/extensions/spatialanalyst/key-features/raster

Take a look at this also: Performing Raster Noise Reduction and Edge Smoothing?

• Although these sound like applicable methods, they do not seem to address this particular question. Could you please explain how you think they would be applied here? Commented Jun 14, 2013 at 14:43

Instead of trying to smooth this thing out I would try and minimize it initially and then smooth it.

Erase the area of the slide from your source DEM and then convert to contours (will have a gap where the slide occurred). now create a TIN so that void area is infilled using a linear interpolation. Convert TIN to DEM. Now subtract this DEM from the DEM with the slide in it and review the resulting "difference" grid. In theory this difference grid should have an average value that is near your 24m. Figure out what the average difference is and use this to guide adjustments to your 24m thickness value. It is obviously not ever going to be perfect, but you should be able to minimize it and then you can do some smoothing. You could also try just running a "Fill Sinks" on the improved version.