I have been tinkering with a network I created and generating service areas. Comparing it to ESRI generated service areas (using ArcGIS Online) for the same point shows a drastically different service area.
So I started looking up network dataset tutorials. Most of these show you how to create a network, but not what a fully functional and accurate one actually has in terms of data (roads, rails, bus stops, stop signs, traffic signals....etc).
So my main question is, what data should I have in my dataset and what would the network settings look like?
My secondary question, is what features and settings does the ESRI network contain?
Just my $00.02....
The ideal network may or may not contain everything you mention above. The first question one needs to ask is
What is my objective when developing my [insert here]?
This goes for just about anything right? Databases, Networks, etc.
I used NA for a number of years for 9-1-1 dispatch. All I needed was a centerline network with speed limits and the locations of fire stations. I used universal turn impedances for right, left, and straight (no turns) through any given intersection. Using speed limits along with centerline segment length, I was able to calculate the amount of time it took to traverse a segment. Some those up and you get a relative drive time to a given incident. The objective was met: I would dispatch based on drive time: Minimum drive time response gets dispatched.
As the complexity of any GIS analysis increases, as does the need for more complex (read; precise and accurate) data. If all you want to do is draw a map of the counties in your state, you just need a polygon feature class of the counties with their names. But what if you want display population distribution, your data needs just changed. Now you want to display the population of a given demographic? You get the picture....
It may be that the settings you have chosen are different than what ESRI has for their network dataset. I have not used ESRI's service, but I wonder if the settings are exposed? That way you could compare them to yours. It could be as simple as different choices for Impedance or Turn Restrictions.
Another potential factor is the data itself. It may be that the data is different. For example, you may have more modern information, where the ESRI data may be missing streets.
Chris Donohue, GISP
As an example of how data may be different, check out the news story I posted in this thread about a roads issue that came up in the City I work at. One of the realities of this business....
EDIT: Scratch that, even though it is listed with a lead in, the story no longer seems viewable (but the ads still work). So no video to watch from the news network.
Chris Donohue, GISP
You can get information on the routing data that is used for the ArcGIS Online services here:
Also note that most commercial data do not include information on which intersections have traffic lights or stop signs.
Hi @JaySandhu you mentioned "Also note that most commercial data do not include information on which intersections have traffic lights or stop signs. "
I'm here after Googling and trying to figure out how to include stop signs/traffic lights into my Network Dataset, and wondering if that is the reason for discrepancies I am seeing in "Esri" drive time vs. real life or Google Maps drive time.
For example this test route below --
Network Analyst says this trip takes 0.94 minutes
However realistically, there are 4 stop signs along this route, making 2 minutes much more accurate, as suggested by Google Maps.
At first I thought my math was wrong but I followed the instructions here
[Length_Field_Feet] / 5280 * 60 / [Speed_Field_MPH]
My team @RyanBohan then suggested potentially the stop signs...
Do you think it is lack of stop signs that is making up the difference? Thanks!
How did you solve to get the "Esri" drivetime of 0.94 Minutes? i.e. Are you using your own network dataset or the Esri Streetmap data hosted online? Because when I solve on that I get 1.7 min when using no set time and 2 minutes when I specify Friday 4 PM. See attached screen shot.
If you are using your own network dataset and calibrating the drive time based on posted speed limit and street length in miles then you will get the "fastest" possible way to drive that path with no traffic, turn penalties or stop sign/red lights taken into account. To get realistic travel times without adding the extra overhead of adding stop sign/traffic light information, you can simply use a slower speed. That is, instead of posted speed limit of 30 miles, you can say that on average cars travel at 20 miles per hour taking into account road conditions. You will get more realistic results. Of course you can add in individual turning movements and slow downs and use traffic counts for the area to come up with better travel times but that is lot more work! Or you can use the Global Turn Category Evaluator to automatically configure some delays at road intersections. You can read more about that here:
Hope this explains some of the nuances of calibrating the road network!
This is really helpful information, thank you for trying it yourself! It makes me hopeful I have something wrong in my network dataset that is fixable.
My screenshot was indeed using my own network, for which building it was an entirely new process for me having only built networks in ArcMap in the past. Unfortunately our org is out of credits and I am unable to use the Online dataset.
You can also download the Network Analyst tutorial data and look at the way the SanFran and SanDiego network datasets have been put together.