Learning How to Use OSMNX: Mapping the Parisian Metro #

September 28, 2023

I’ve been recently toying around with osmnx quite a bit, mostly because I really wanted to do some urban design and city modeling. I discovered osmnx through this and later, through Geoff Boeing’s Blog, became extremely excited at the idea.

So I figured I may as well give it a go. I haven’t done or worked on many personal projects in a while so this did excite me quite a lot.

Firstly, I had to choose a city and an idea that I’d model out. Something pretty basic to start with in my mind would be modeling a city itself, and then perhaps do some kind of shortest path earch on the metro overlayed upon the city.

Because I have an immense bias to everything French, I immediately chose to do Paris.

It’s pretty easy to first pull up a city and model it through a network with osmnx:

G = ox.graph_from_place("Paris, France")
ox.plot_graph(G)

The data we recieve here is all modeled with networkx, so it is, in effect, a modeling of nodes and edges. In this case, I believe we are modeling the streets/roads of Paris and the connections between them. I’m not a huge fan of how this looks, so let’s go ahead and change it up a bit.

Let’s start by taking a look at our data itself, and how it’s structured:

gdf_nodes, gdf_edges = ox.graph_to_gdfs(G)
gdf_nodes.head()

We can take this data and just restructure it and redesign its output and thus its appearance:

G2 = ox.graph_from_gdfs(gdf_nodes, gdf_edges, graph_attrs=G.graph)
ox.plot_graph(G2, node_size=0, bgcolor="#ffffff", node_color='b')

Alright, now that we have a somewhat decent model of Paris up, we can now consider their metro system. We can do this by performing a filter on the data that we get from osmnx where we look in particular for railways and subways (["railway"~"subway"]).

G3 = ox.graph_from_place("Paris, France", retain_all=False, truncate_by_edge=True, simplify=True,                        custom_filter='["railway"~"subway"]')
ox.plot_graph(G3)

Alright! So that looks about right. The design isn’t fantastic, but we can fix that, or we can make it look even more ridiculous! Let’s go ahead and pull out all of the information from this metro data and overlay it onto the city data we got from earlier. We can redesign it as we do that as well.

composite_G = nx.compose(G2, G3)
ec = ['y' if 'Métro' in str(c) else '#8a3535' for _,_,c in composite_G.edges.data('name')]
fig, ax = ox.plot_graph(composite_G, bgcolor='k', edge_color=ec, node_size=0, 
                        edge_linewidth=0.5, close=False)

Now, lets take a look at our data and see what we can do with it. I’d ideally like to do a shortest path travelersal on some of the stations of the Paris metro, because this is all built on networkx, which already provides that functionality, it shouldn’t be too difficult.

G3.edges.data()

OutMultiEdgeDataView([(29588070, 260767428, {'osmid': 142160901, 'tunnel': 'yes', 'name': 'Métro 9', 'oneway': False, 'reversed': False, 'length': 1977.8989999999997, 'geometry': <LINESTRING (2.244 48.834, 2.256 48.838, 2.257 48.838, 2.257 48.838, 2.258 4...>}), (134299001, 10038858874, {'osmid': 24331466, 'tunnel': 'yes', 'name': 'Métro 12', 'oneway': False, 'reversed': False, 'length': 1631.086, 'geometry': <LINESTRING (2.273 48.824, 2.274 48.825, 2.275 48.825, 2.276 48.825, 2.277 4...>}),

Hmmm… There doesn’t really appear to be much information there about the name of the stations themseleves. Let’s see if we can dig deeper:

dat = [c for _,_,c in G3.edges.data()]
for i in range(0, len(dat)):
    print(dat[i]['name'])

Métro 9
Métro 12
Métro 6 Corvisart-Place D'Italie
Métro 6
Métro 2
Métro 2
...

Ah. So only a select few stations are actually labeled by their proper names, the rest appear to just be tagged to their line? That is quite unfortunate. In that case, we can at least go off of the osmid, which looks like a unique identifier for these different transportation nodes, but I can’t say for sure. Let’s take, for instance 29588070 as our beginning station on the metro and 325664764 as our destination station. I’ll use the Dijkstra’s method for this.

path = nx.shortest_path(G3, source=29588070, target=325664764, method='dijkstra')
nc = ['#025228' if a in path else '#9aa99a' for a,_,_ in composite_G.edges.data()]
el = [1 if a in path else 0.5 for a,_,_ in composite_G.edges.data()]
fig, ax = ox.plot_graph(composite_G, bgcolor='#ffffff', edge_color=nc, node_size=0, 
                        edge_linewidth=el, close=False)

Interesting, looks like we managed to get some stop at the edge of one of the lines leading to what looks like a stop somewhere in the North (or in the oposite direction). Pretty neat! Unfortunately, it would be a bit difficult to build a system out of this that would let me choose two different stations by name and then see the shortest path that way, but perhaps another city would have those options. For now, this was at least a good enough introduction for me into osmnx. I really like all the features available with this, and I’ll be trying to do a lot more with it in the future.