Cartographically is there any reason why (pro / con) multi-part line features should not be used when developing road networks for ArcGIS Network Analyst?

For example, say I have a road which is 100 units long and it has two bridges; one located at 20-25 units and one located at 50-55 units. The road is one continuous line feature with no intersecting roads.

So is it better to have one multi-part feature (one database entry) comprised of three features (segment before bridge one, the segment between bridges one and two, and the segment after bridge two), or three feature records in the feature class – one record for each road segment and not have a multi-part feature? Multi-part Road Example

1 Answer 1


The road goes over the bridge. In either scenario you describe, the network will fail/have gaps there with no connectivity.

If the bridge is important, those segments should still be present in the road layer but as their own separate features. In such a case (really any case) you would not want a multipart feature on either end and in between. You either want a single part feature that runs over both bridges (whether that line has multiple vertices or just start/end nodes is up to you) or you want five single part features.

It somewhat depends on how and what your network model is going to do (ie, weight limits for particular edges vs using line barriers snapped to the edge where a bridge is, or just treating the entire edge as a single entity with restrictions based on the parts). I would avoid multipart features unless you have a specific need or reason for them. Within the network they will be considered invalid geometries because they break connectivity.

  • ![multi-part example](C:\Users\Jeff Reichman\Pictures\Untitled.png)
    – Jeff
    Apr 1, 2015 at 0:20
  • Chris thanks for the comments. I uploaded an example image. Highlighted in blue is the multi-part line and the bridges are in purple. SO you really think I should break my road (in this case) into four separate records.
    – Jeff
    Apr 1, 2015 at 0:27
  • So there is no road feature coincident with the bridges? If that is the case, then my first sentence still applies. When you go to turn the blue road lines into a network, the gaps where the bridges are will result in no connectivity (and two disconnected edges). In this case you actually need seven features if you want the bridge segments to have different attributes than the others, or one single part blue line that just goes right over and all the way through the bridges (ideally coincident, meaning snapped/shared vertices). The road features must be continuous for connectivity.
    – Chris W
    Apr 1, 2015 at 2:12
  • The bridges do snap to the road features. I think the whole matter is some of my folks aren't use to working with or have multipart features. I suppose I can just run the "Multipart To Singlepart" and make everyone happy. Appreciate the feedback.
    – Jeff
    Apr 1, 2015 at 2:20
  • 1
    @JeffReichman snapping endpoints is only part of 'coincident'. Full coincidence means there's two overlapping lines. Basically when you go to create a network, you need to look at a single polyline layer (roads). Any gaps in that will be gaps in the network. Other feature classes are either a) irrelevant to the network (like bridges - those are handled via edge attributes and barriers and such within a network) or b) turn it into a multi-modal network (think train lines and roads, or ferry lines in water and roads, or all three).
    – Chris W
    Apr 1, 2015 at 2:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.