# network based interpolation methods, which one could I use?

I need to interpolate traffic data consisting in raw number of vehicles passing across certain points, where counters are. It is evident I haven't a uniform distribution of points and values about traffic. I would like to ask which methods I could use to interpolate this network deducing traffic for each location of the overall network. I tried to look on the most used desktop GIS (QGIS, ArcGIS) but it seems there is not a'ready made' solution at the moment. Is there an algorithm that I can implement in Python to achieve this result? Thanks!

• This doesn't sound like a network analysis, but rather a spatial one. You have point sample values, which could be interpolated to form a 'traffic level contour' map. You could then intersect that layer with the network to assign a traffic factor to each edge. But this would be a generalization at best given the nature of a network. I have seen some more specific analysis done along this line of thinking as related to emergency evacuation routes. You may find some more information searching on that subject. – Chris W May 18 '14 at 19:10
• What you say can be considered a solution I have already thought about but the thing is that vehicles do not move in a euclidean space and most of the times road circulation has barriers, as one-way roads and not allowed turns. I've found an interesting paper on the International Journal of Geographical Information Science Vol. 26, No. 4, April 2012, 667–689 in which they used a network based metric (modifying euclidean distances) using universal kriging. My maths competencies are not that good and if there is a software doing the job that would be great – user3523583 May 19 '14 at 16:41
• I should say spatial constrained to network rather than not network at all. Restrictions are a factor, which is why I said generalization. Just considering a main vs side street there would be a large difference in traffic not reflected by that method. It would be more accurate with classed roads and only using point samples on the same class. I did find this paper suggesting minimal improvement with network distances over Euclidean. Have you looked at TransCAD or similar transportation planning software? – Chris W May 19 '14 at 18:49