v.clean (GRASS) creates additional features

Question

I am working with a large road dataset (approx. 1 million features) with a high number of overshoots and dangles.

To address the overshoots and dangles, I am using the v.clean tools (break, snap, rmdangle) with the -c flag to combine with follow-up tools. I have also selected only lines as the input feature type and added threshold values of 1 and 7 (ft) for snap and rmdangle.

After running this tool, the number of features increases by approximately 40,000. I am looking for clarity on the cause of these additional features.

I am under the impression that the break tool is primarily responsible for creating the extra features. What I am not clear on is exactly what format these extra features are in and why they are being created. I am assuming each segment of the break becomes a new feature? But in that case, how are original feature attributes handled? For instance, I have a segment length attribute and a travel time attribute, and it is important that the integrity of those attributes remain in the cleaned dataset. In addition, I have noticed that changing the threshold value on snap also changes the number of features, and I am unclear why that would be.

In addition, from the GRASS GIS manual, I have noted that breaking lines should be followed by removing duplicates. Since I have added the -c flag, it seems that duplicates should be removed. Could the extra features be duplicates?

Answer 1

I think that your impression is correct. Often non-topological line datasets will have lines crossing with no node at that intersection. v.clean with the break tool will split the crossed line, thus creating additional line segments. So you should update the segment lengths after doing the v.clean.

Furthermore changing the threshold will, of course, change which dangles are removed, and which are left. You should add a threshold for each tool, even though break does not need a threshold. So your thresh value should be three numbers: "0,1,7" (or whatever you choose).