How would I find areas in a dem representing say at least a 5 foot change in elevation within a maximum horizontal distance of 30 feet?

Question

I have attempted doing a slope analysis of a DEM and isolated areas of interest, however I only want areas where the elevation changes a set distance within at least 30 horizontal feet.  The slope tool gives me down to individual 1x1 meter areas representing that slope (e.g. 16.67%;  5 ft rise over 30 ft run).  Is there a way to identify areas where such an elevation changes over a set distance horizontally?

DanPatterson · Answer

You could use a moving or block window of using a 30 ft cell size,

Focal Statistics (Spatial Analyst)—ArcGIS Pro | Documentation

Block Statistics (Spatial Analyst)—ArcGIS Pro | Documentation

the max minus the min will give you the vertical distance relative to the horizontal minimum.

This is your "mask" raster which delineating those areas that meet your threshold.

A grouping of the above zones using Region Group (Spatial Analyst)—ArcGIS Pro | Documentation

will give you individually labelled regions that you can analyze for size and/or shape characteristics using

Zonal Geometry (Spatial Analyst)—ArcGIS Pro | Documentation

or a variety of other tools in the Spatial Analyst toolset (eg Surface Parameters (Spatial Analyst)—ArcGIS Pro | Documentation)

DanBennett · Answer

Thank you for the reply. I have essentially gotten to that point, my main trouble is only selecting areas where that elevation change only occurs over at least 30 feet of horizontal distance. After converting polygon, I attempted to calculate an average width of each area using Area and Perimeter and then removing those that fall below that threshold. Was thinking a cleaner way might exist.

quillaja · Answer

Would using the "Reclassify" tool with ranges such as 0-16.67% => 1, 16.671-MAX% => 0 (or nodata) do what you want? This would select or group (however you want to think about it) all the areas with slope <= 16.67% into the "1" category and all other areas into "0" or nodata.