root676
commented
5 years ago Thank you for pointing this out - I will still mark this as wontfix because it is the users responsibility to prepare his/her datasets in a reasonable manner (in the same way as with adding reasonable cost values or oneway information). This case is certainly not a common one and therefore has to be preprocessed by the user. Furthermore, adding the parameters for a user side densification would result in unnecessarily complex algorithm input UIs (...and they are already pretty complex).
In order to get more accurate Iso-Area results, users may need to densify their Network Layer input using the Densify by interval algorithm.
I tested the algorithm by constructing a few very long lines with few vertices and calculating Iso-Areas at intervals that would fall in between distant vertices. Since the algorithm first calculates a point cloud using all the vertices in the network and does not create any new points , it means that the outer-most iso-area contour or polygon may underestimate cost by as much as the longest segment of a line.
Most network inputs will already have reasonably dense vertices: this is only a problem if you have some very long, straight line segments.
E.g. see below, the 2500m Iso-Area point cloud and interpolation originating at the red hospital icon:
and the same result after using Densify network:
(I'm not sure what happened to the interpolation at the southern end of this figure... there are rows of nodata, but nonetheless you can see the difference in the point cloud)