Buffer Results not logical

Question

I am buffering a large polyline dataset (streets).  Very often when I run the buffer I get large polygons that are created as if the buffer was large enough between the two lines to create one polygon.  The buffer, however, is less than the distance it would take to close off the polygon.  The picture will make more sense.

The red is the original polyline.  Notice the large closed off polygons.  It seems like there are times when it runs without this issue.

Any thoughts?

bufferpolyerror.JPG

ChrisSnyder · Answer

True curves are evil indeed. I have had all sorts of issues with "true curve" features and many of the analysis tools. As memory serves, the buffer tool won't output true curves if you don't use the "dissolve" option...

Work around: Shapefiles don't support true curves :) . Sorry to say, but shapefiles are still generally faster than FGDB and certainly PGDB...

JimW1 · Answer

It is projected and buffer is in the same units.  It is a very large dataset so I had it in a model that clipped it first then buffered.  I took the intermediate clipped results and buffered them separately and it seems to have worked.  But i seem to remember it failed once before even when I did it this way.  Anyhow, I got the result I wanted and dont have time isoloate the root problem.

Thanks for you input.

My experience is that when this happens it is often due to ArcGIS mishandling polylines that are curves. I learned the hard way not to buffer a buffer because the buffers on sharp edges are a two-point curve as stored in a geodatabase. My guess with your data is that somewhere there is a very sharp angled street being buffered and the resulting buffer is crossing over itself and then filling in the other areas.

Also, line segments that cross themselves will also do some odd things.

My solution was to first generalize the buffers then re-buffer. The quicker solution was to use shapefiles but that didn't fit my workflow.

RustyRex · Answer

It is projected and buffer is in the same units.  It is a very large dataset so I had it in a model that clipped it first then buffered.  I took the intermediate clipped results and buffered them separately and it seems to have worked.  But i seem to remember it failed once before even when I did it this way.  Anyhow, I got the result I wanted and dont have time isoloate the root problem.

Thanks for you input.