I have GPS data (X,Y, speed, CO2 and fuel consumption) for one truck for one trip. I want to import them in ArcGIS and to join them with the road network (map matching). I was thinking of applying "near" to the data points so that they can get road segments' id and then "summarize" to get data for the entire road segment and join them afterwards with the road network.

My question is "near" reliable enough? For CO2 emissions if i have two points recorded in a road segment should i take the total or the average?


1 Answer 1


All sounds sensible to me but depending on what your overall aim is you might want to split long segments into shorter ones of equal length, as links in road network can be anywhere from 2 metres to 10 miles long so your spatial resolution will vary enormously (unless it is the link itself you are interested in, which for a pollution problem seems unlikely). I think there is an Arc tool for this though can't find it right now.

As to mean or total, that depends on your end use again. Do you have enough traces to reliably estimate traffic levels per link? Then you could sum. If not then take mean though bear in mind this will introduce heteroscedasticity.

As to whether 'near' is reliable, the only pitfall I see is erroneous GPS points that either (1) don't fall near any link, (2) fall close to an intersection so it's not possible to tell which link they are on, or (3) are recorded in built up areas with poor signal and appear to fall on a neighbouring link instead. (1) and (2) can easily be discarded if you can afford to lose the data. (3) is more insidious and can't be solved without far more advanced data cleaning, but you may just decide it's not a problem.

  • 1) I want to create link-emissions maps and 2) i want to compare link emissions from real world (GPS) data to emissions estimated in an emission model. The estimated emissions refer to passenger cars as a total while i have gps data from 50 passenger cars. i was thinking of taking the mean and compare it with estimated emissions' mean. How do you think i should do it? Should i split long segments? Nov 16, 2015 at 14:47
  • If your emissions model predicts identical emissions along the length of each link then that would be a good reason not to split into shorter segments. Otherwise, split to approximately match the spatial resolution of the emissions model. What is your emissions model? Nov 17, 2015 at 15:50
  • I use COPERT. I believe it's not necessary to split them. But how do you think could i bring emissions in a same level of comparison? Should i get the average and then multiply it by total number of flows given for a specific link? or just compare average emissions from emission model and GPS data? Nov 18, 2015 at 8:13
  • Does COPERT output emissions levels per vehicle, or per link? Presumably per vehicle would match a mean and per link would match a sum. If you have the choice of either, then as I said above, be guided by your own aims. Either way, don't try comparing on links that have no GPS traces. Nov 18, 2015 at 9:01
  • It outputs emissions per vehicle type and for different road classifications. So, i then assign them to my road network depending on highway-urban-arterial segments. The output must be a per-link comparison not total of vehicles. So,you suggest of summing the records and then multiply them by the total volume on the link? In practice i want to examine how does the model react to reality, so if the links (with GPS traces) agree to an extent then i can assume that the model makes a good estimation for the rest of the network as well. Nov 18, 2015 at 10:54

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.