RivEX

The network shown below was derived from a DEM for the SNAPP Amazon project. The result is a network comprised of 4,083,180 polylines with an unusually large number of pseudo nodes (187,110 in total).  Experience has shown that raster derived network can often have an unusually high number of pseudo nodes and thus benefit by the removal of such nodes.

The Amazon basin, SNAPP Amazon project.Stream orders 1 to 3 have been removed for clarity

Below is a zoomed in section of the network showing pseudo nodes as black points.

Such a network can cause recursive algorithms to reach such depths that they cause a stack overflow error. This can be resolved by using RivEX to remove pseudo nodes. This simplifies the topology of the network and can significantly reduce the number of features in the dataset without loss of connectivity or change in geometry. The Amazon network shown above was processed by RivEX to remove pseudo nodes resulting in a new network comprised of 3,895,952 polylines, that's 187,228 less features the topology needs to record, with no loss to connectivity or change in geometry!  In this example this may have only reduced the total number of features in this network by 4.5% but in other networks it can be considerably more (e.g. > 90% reduction).

Such a massive reduction in number of polylines means the topological dictionaries created by RivEX are significantly smaller, load faster, process faster and will not encounter stack overflow issues.