Archive

PostGIS

This post covers a simple approach to calculating isochrones in a public transport network using pgRouting and QGIS.

For this example, I’m using the public transport network of Vienna which is loaded into a pgRouting-enable database as network.publictransport. To create the routable network run:

select pgr_createTopology('network.publictransport', 0.0005, 'geom', 'id');

Note that the tolerance parameter 0.0005 (units are degrees) controls how far link start and end points can be apart and still be considered as the same topological network node.

To create a view with the network nodes run:

create or replace view network.publictransport_nodes as
select id, st_centroid(st_collect(pt)) as geom
from (
	(select source as id, st_startpoint(geom) as pt
	from network.publictransport
	) 
union
	(select target as id, st_endpoint(geom) as pt
	from network.publictransport
	) 
) as foo
group by id;

To calculate isochrones, we need a cost attribute for our network links. To calculate travel times for each link, I used speed averages: 15 km/h for buses and trams and 32km/h for metro lines (similar to data published by the city of Vienna).

alter table network.publictransport add column length_m integer;
update network.publictransport set length_m = st_length(st_transform(geom,31287));

alter table network.publictransport add column traveltime_min double precision;
update network.publictransport set traveltime_min = length_m  / 15000.0 * 60; -- average is 15 km/h
update network.publictransport set traveltime_min = length_m  / 32000.0 * 60 where "LTYP" = '4'; -- average metro is 32 km/h

That’s all the preparations we need. Next, we can already calculate our isochrone data using pgr_drivingdistance, e.g. for network node #1:

create or replace view network.temp as
 SELECT seq, id1 AS node, id2 AS edge, cost, geom
  FROM pgr_drivingdistance(
    'SELECT id, source, target, traveltime_min as cost FROM network.publictransport',
    1, 100000, false, false
  ) as di
  JOIN network.publictransport_nodes pt
  ON di.id1 = pt.id;

The resulting view contains all network nodes which are reachable within 100,000 cost units (which are minutes in our case).

Let’s load the view into QGIS to visualize the isochrones:

isochrone_publictransport_1

The trick is to use data-defined size to calculate the different walking circles around the public transport stops. For example, we can set up 10 minute isochrones which take into account how much time was used to travel by pubic transport and show how far we can get by walking in the time that is left:

1. We want to scale the circle radius to reflect the remaining time left to walk. Therefore, enable Scale diameter in Advanced | Size scale field:

scale_diameter

2. In the Simple marker properties change size units to Map units.
3. Go to data defined properties to set up the dynamic circle size.

datadefined_size

The expression makes sure that only nodes reachable within 10 minutes are displayed. Then it calculates the remaining time (10-"cost") and assumes that we can walk 100 meters per minute which is left. It additionally multiplies by 2 since we are scaling the diameter instead of the radius.

To calculate isochrones for different start nodes, we simply update the definition of the view network.temp.

While this approach certainly has it’s limitations, it’s a good place to start learning how to create isochrones. A better solution should take into account that it takes time to change between different lines. While preparing the network, more care should to be taken to ensure that possible exchange nodes are modeled correctly. Some network links might only be usable in one direction. Not to mention that there are time tables which could be accounted for ;)

Advertisement

This post is a quick instruction for installing Postgres 9.2, PostGIS 2.0 and pgRouting 2.0.

  1. For Postgres, download the installer from enterprisedb.com.
  2. Run the installer. You’ll have to pick a superuser password – remember it, you’ll need it again soon.
  3. At the end of the installation process, allow the installer to start Stack Builder.
  4. In Stack Builder, select the Postgres 9.2 installation and install PostGIS from the list of available extensions.
  5. The PostGIS installation procedure will prompt you for the superuser password you picked before.
  6. I suggest letting the installer create a sample database We’ll need it later anyway.

Now for the pgRouting part:

  1. Download the pgRouting zip file for your system (32 or 64 bit) from Winnie.
  2. Unzip the file. It contains bin, lib and share folders as well as two text files.
  3. Copy these folders and files over to your Postgres installation. In my case C:\Program Files\PostgreSQL\9.2

Installation – done.

Next, fire up pgAdmin. If you allowed the PostGIS installer to create a sample database, you will find it named postgis20 or similar. Otherwise just create a new database using the PostGIS template database. You can enable pgRouting in a database using the following steps:

  1. In postgis20, execute the following query to create the pgrouting extension. This will add the pgRouting-specific functions:
    CREATE EXTENSION pgrouting;
  2. Test if everything worked fine:
    SELECT pgr_version();

    It should return "(2.0.0-dev,v2.0.0-beta,18,a3be38b,develop,1.46.1)" or similar – depending on the version you downloaded.

pgadmin_pgrouting

How about some test data? I’ll be using the public transport network of the city of Vienna provided in GeoJSON format from http://data.wien.gv.at/daten/geo?service=WFS&request=GetFeature&version=1.1.0&typeName=ogdwien:OEFFLINIENOGD&srsName=EPSG:4326&outputFormat=json:

    1. Just copy paste the url in Add Vector Layer | Protocol to load the dataset.
    2. Use DB Manager to load the layer into your database. (As you can see in the screenshot, I created a schema called network but that’s optional.)

import_publictransport

  1. To make the line vector table routable, we use pgr_createTopology. This function assumes the columsn called “source” and “target” exist so we have to create those as well:
    alter table network.publictransport add column source integer;
    alter table network.publictransport add column target integer;
    select pgr_createTopology('network.publictransport', 0.0001, 'geom', 'id');
    

    I’m quite generously using a tolerance of 0.0001 degrees to build the topology. Depending on your dataset, you might want to be more strict here.

  2. Let’s test it! Route from source #1 to target #3000 using pgr_dijkstra:
    SELECT seq, id1 AS node, id2 AS edge, cost, geom
      FROM pgr_dijkstra(
        'SELECT id, source, target, st_length(geom) as cost FROM network.publictransport',
        1, 3000, false, false
      ) as di
      JOIN network.publictransport pt
      ON di.id2 = pt.id ;

    Note how the query joins the routing results and the network table together. (I’m aware that using the link length as a cost attribute will not lead to realistic results in a public transport network but bear with me for this example.)

  3. We can immediately see the routing results using the Load as new layer option:

select_dijkstra
route

Nice work! pgRouting 2.0 has come a long way. In a post from April this year, Boston GIS even announced to add pgRouting into the Stack Builder. That’s going to make the installation even more smooth.

This is an update to my previous post “WFS to PostGIS in 3 Steps”. Thanks to Even Rouault’s comments and improvements to GDAL, it is now possible to load Latin1-encoded WFS (like the one by data.wien.gv.at) into PostGIS in just one simple step.

To use the following instructions, you’ll have to get the latest GDAL (release-1600-gdal-mapserver.zip)

You only need to run SDKShell.bat to set up the environment and ogr2ogr is ready for action:

C:\Users\Anita>cd C:\release-1600-gdal-mapserver
C:\release-1600-gdal-mapserver>SDKShell.bat
C:\release-1600-gdal-mapserver>ogr2ogr -overwrite -f PostgreSQL PG:"user=myuser password=mypassword dbname=wien_ogd" "WFS:http://data.wien.gv.at/daten/geoserver/ows?service=WFS&request=GetFeature&version=1.1.0&typeName=ogdwien:BEZIRKSGRENZEOGD&srsName=EPSG:4326"

Thanks everyone for your comments and help!

This is a quick note on how to download features from a WFS and import them into a PostGIS database. The first line downloads a zipped Shapefile from the WFS. The second one unzips it and the last one loads the data into my existing “gis_experimental” database:

wget "http://data.wien.gv.at/daten/geoserver/ows?service=WFS&request=GetFeature&version=1.1.0&typeName=ogdwien:BEZIRKSGRENZEOGD&srsName=EPSG:4326&outputFormat=shape-zip" -O BEZIRKSGRENZEOGD.zip
unzip -d /tmp BEZIRKSGRENZEOGD.zip
shp2pgsql -s 4326 -I -S -c -W Latin1 "/tmp/BEZIRKSGRENZEOGD.shp" | psql gis_experimental

Now, I’d just need a loop through the WFS Capabilities to automatically fetch all offered layers … Ideas anyone?

Thanks to Tim for his post “Batch importing shapefiles into PostGIS” which was very useful here.

Update: Many readers have pointed out that ogr2ogr is a great tool for this kind of use cases and can do the above in one line. That’s true – if it works. Unfortunately, it is picky about the supported encodings, e.g. doesn’t want to parse ISO-8859-15. In such cases, the three code lines above can be a good alternative.

This is the follow up post to “An osm2po Quickstart” which covers loading the OSM network into PostGIS and using the result with pgRouting. After parsing the OSM file, e.g.

C:\Users\Anita\temp\osm2po-4.2.30>java -jar osm2po-core-4.2.30-signed.jar prefix=at "C:\Users\Anita\Geodaten\OpenStreetMap Data\austria.osm.pbf"

you should find a folder with the name of the prefix you chose inside the osm2po folder. It contains a log file which in turn provides a command line template for importing the OSM network into PostGIS, e.g.

INFO commandline template: psql -U [username] -d [dbname] -q -f "C:\Users\Anita\temp\osm2po-4.2.30\at\at_2po_4pgr.sql"

Using this template, we can easily import the .sql file into an exiting database. My pgRouting-enabled database is called wien_ogd.

C:\Users\Anita\temp\osm2po-4.2.30\at>psql -U [username] -d wien_ogd -q -f C:\Users\Anita\temp\osm2po-4.2.30\at\at_2po_4pgr.sql

Now, the data is ready for usage in QGIS:

The osm2po table in QGIS

Using “pgRouting Layer” plugin, it’s now straightforward to calculate shortest paths. I had to apply some changes to the plugin code, so please get the latest version from Github.

A shortest path in osm2po network

Using osm2po turned out to be far less painful than I expected and I hope you’ll find this post useful too.

The function with the glorious name “find_node_by_nearest_link_within_distance” is part of pgRouting and can be found in matching.sql.

“This function finds nearest node as a source or target of the nearest link”
That means that we can use this function e.g. to find the best road network node for a given address.

The function returns an object of type link_point:

CREATE TYPE link_point AS (id integer, name varchar);

To access only the id value of the nearest node, you can use:

SELECT id(foo.x) 
FROM (
   SELECT find_node_by_nearest_link_within_distance(
	'POINT(14.111 47.911)',
	0.5,
	'nw_table')::link_point as x
) AS foo

Alpha shapes are generalizations of the convex hull [1]. Convex hulls are well known and widely implemented in GIS systems. Alpha shapes are different in that they capture the shape of a point set. You can watch a great demo of how alpha shapes work on François Bélair’s website “Everything You Always Wanted to Know About Alpha Shapes But Were Afraid to Ask” I borrowed the following pictures from that site:

Alpha shapes for different values of alpha. The left one equals the convex hull of the point set.

Alpha shapes for different values of alpha. The left one equals the convex hull of the point set.

pgRouting comes with an implementation of alpha shapes. There is an alpha shape function: alphashape(sql text) and a convenience wrapper: points_as_polygon(query character varying). The weird thing is that you don’t get to set an alpha value. The only thing supplied to the function is a set of points. Let’s see what kind of results it produces!

Starting point for this experiment is a 10 km catchment zone around node #2699 in my osm road network. Travel costs to nodes are calculated using driving_distance() function. (You can find more information on using this function in Catchment Areas with pgRouting driving_distance().)

CREATE TABLE home_catchment10km AS
SELECT *
   FROM osm_nodes
   JOIN
   (SELECT * FROM driving_distance('
      SELECT gid AS id,
          source,
          target,
          meters AS cost
      FROM osm_roads',
      2699,
      10000,
      false,
      false)) AS route
   ON
   osm_nodes.id = route.vertex_id

After costs are calculated, we can create some alpha shapes. The following queries create the table and insert an alpha shape for all points with a cost of less than 1500:

CREATE TABLE home_isodist (id serial, max_cost double precision);
SELECT AddGeometryColumn('home_isodist','the_geom',4326,'POLYGON',2);

INSERT INTO home_isodist (max_cost, the_geom) (
SELECT 1500, ST_SetSRID(the_geom,4326)
FROM 
  points_as_polygon(
    'SELECT id, ST_X(the_geom) AS x, ST_Y(the_geom) AS y FROM home_catchment10km where cost < 1500'));

In previous posts, I’ve created catchment areas by first interpolating a cost raster and creating contours from there. Now, let’s see how the two different approaches compare!

The following picture shows resulting catchment areas for 500, 1000, 1500, and 2000 meters around a central node. Colored areas show the form of pgRouting alpha shape results. Black contours show the results of the interpolation method:

Comparison of pgRouting alpha shapes and interpolation method

At first glance, results look similar enough. Alpha shape results look like a generalized version of interpolation results. I guess that it would be possible to get even closer if the alpha value could be set to a smaller value. The function should then produce a finer, more detailed polygon.

For a general overview about which areas of a network are reachable within certain costs, pgRouting alpha shapes function seems a viable alternative to the interpolation method presented in previous posts. However, the alpha value used by pgRouting seems too big to produce detailed catchment areas.

[1] http://biogeometry.duke.edu/software/alphashapes/

Advanced labeling in QGIS new labeling engine is mostly about data-defined settings. Almost any property of the label can be controlled.

For this example, we will try to mimic the look of the classic Google map with it’s line and label styles. The data for this post is from the OpenStreetMap project provided as Shapefiles by Cloudmade.

After importing the roads into PostGIS using PostGIS Manager Plugin, we can create a view that will contain the necessary label style information. The trick here is to use CASE statements to distinguish between different label “classes”. Motorway labels will be bigger than the rest and the buffer color will be the same color as used for the corresponding lines.

DROP VIEW IF EXISTS v_osm_roads_styled;

CREATE VIEW v_osm_roads_styled AS
SELECT *, 
CASE WHEN type = 'motorway' THEN 9
     ELSE 8 END
     as font_size,
'black'::TEXT as font_color,
false as font_bold,
false as font_italic,
false as font_underline,
false as font_strikeout,
false as font_family,
1 as buffer_size,
CASE WHEN type = 'motorway' THEN '#fb9139'::TEXT
     WHEN type IN ( 'primary','primary_link','secondary','secondary_link') THEN '#fffb8b'::TEXT
     ELSE 'white'::TEXT END 
     as buffer_color
FROM osm_roads;

In QGIS, we can then load the view and start styling. First, let’s get the line style ready. Using rule-based renderer, it’s easy to create complex styles. In this case, I’ve left it rather simple and don’t distinguish between different zoom levels. That’s a topic for another post :)

Google-style rules for OSM road data

Now for the labels! In “Data defined settings”, we can assign the special attributes created in the database view to the settings.

Completed "Data defined settings"

To achieve an even better look, go to “Advanced” tab and enable “curved” and “on line” placement. “Merge connected lines to avoid duplicate labels” option is very helpful too.

Finally – after adding some water objects (Cloudmade natural.shp) – this is what our result looks like:

Google-style OSM map

This solution can be improved considerably by adding multiple zoom levels with corresponding styles. One obvious difference between the original Google map and this look-alike is the lack of road numbers. Tim’s post on “shield labels” can be a starting point for adding road numbers the way Google does.

Today, I’ve been experimenting with data from OpenFlights.org. They offer airport, airline and route data for download. The first idea that came to mind was to connect airports on a shared route by lines. This kind of visualization just looks much nicer if the connections are curved instead of simple straight lines.

Luckily, that’s pretty easy to do using PostGIS. After loading airport positions and route data, we can create the connection lines like this (based on [postgis-users] Great circle as a linestring):

UPDATE experimental.airroutes
SET the_geom = 
(SELECT ST_Transform(ST_Segmentize(ST_MakeLine(
       ST_Transform(a.the_geom, 953027),
       ST_Transform(b.the_geom, 953027)
     ), 100000 ), 4326 ) 
 FROM experimental.airports a, experimental.airports b
 WHERE a.id = airroutes.source_id  
   AND b.id = airroutes.dest_id
);

The CRS used in the query is not available in PostGIS by default. You can add it like this (source: spatialreference.org):

INSERT into spatial_ref_sys (srid, auth_name, auth_srid, proj4text, srtext) values ( 953027, 'esri', 53027, '+proj=eqdc +lat_0=0 +lon_0=0 +lat_1=60 +lat_2=60 +x_0=0 +y_0=0 +a=6371000 +b=6371000 +units=m +no_defs ', 'PROJCS["Sphere_Equidistant_Conic",GEOGCS["GCS_Sphere",DATUM["Not_specified_based_on_Authalic_Sphere",SPHEROID["Sphere",6371000,0]],PRIMEM["Greenwich",0],UNIT["Degree",0.017453292519943295]],PROJECTION["Equidistant_Conic"],PARAMETER["False_Easting",0],PARAMETER["False_Northing",0],PARAMETER["Central_Meridian",0],PARAMETER["Standard_Parallel_1",60],PARAMETER["Standard_Parallel_2",60],PARAMETER["Latitude_Of_Origin",0],UNIT["Meter",1],AUTHORITY["EPSG","53027"]]');

This is an example visualization (done in QGIS) showing only flight routes starting from Vienna International Airport:

Flight routes from Vienna International

Connections crossing the date line are currently more problematic. Lines would have to be split, otherwise this is what you’ll get:

Date line trouble

Do you need to create a table with a geometry column in PostGIS from scratch?
Can’t remember the syntax of AddGeometryColumn(varchar catalog_name, varchar schema_name, varchar table_name, varchar column_name, integer srid, varchar type, integer dimension)? I can’t. ;)

Let’s make our lives easier: QGIS PostGIS Manager offers a convenient GUI for creating tables with geometry columns:

PostGIS Manager Create Table dialog

The dialog works a lot like what you’re probably used to in pgAdmin – with the added nicety of supporting Geometry columns.

In my opinion, this is the fastest way so far to create a spatially enabled table. It provides a much better user experience than telling users (especially new ones) to use AddGeometryColumn(…

%d bloggers like this: