graph_build

Graph build

Graph format and attributes

For a graph file to be usable in Green Paths, it needs to be in the GraphML file format and feature at least the required edge and node attributes listed in the tables below. All possible edge and node attributes are defined (and described) in common/igraph.py.

Node attributes

Alias	Name	Type	Nullable	Description
geometry	geom	Shapely point geometry / WKT	no	Location of the node as a Shapely Point geometry (e.g. POINT (25498747.97 6676644.78)).

Edge attributes

Alias	Name	Type	Nullable	Nodata value	Description
id_ig	ii	int	no		Unique ID of the edge
id_way	iw	string	no		ID of the geometry of the edge (not all geometries are unique)
uv	uv	tuple	no		IDs of the start and end nodes of the edge (e.g. (2046, 4576))
geometry	geom	Shapely LineString geometry / WKT	yes	GEOMETRYCOLLECTION EMPTY	Projected geometry of the edge (e.g. LINESTRING (25498199 6677347, 25498191 6677339)). GP uses EPSG:3879 by default.
geom_wgs	geom_wgs	Shapely LineString geometry / WKT	yes	GEOMETRYCOLLECTION EMPTY	Geometry of the edge in WGS (EPSG:4326) coordinates (e.g. LINESTRING (24.967528 60.208895, 24.967381 60.208826)).
length	l	float	no		Length of the edge in meters.
is_stairs	b_st	boolean	no		A boolean variable indicating whether the edge represents stairs.
allows_biking	b_aw	boolean	no		A boolean variable indicating whether the edge allows biking.
bike_safety_factor	bsf	float	no		Biking safety factor of the edge (calculated by OTP based on OpenStreetMap tags of the edge)
noises	n	dictionary	yes	None	Exposures to different noise levels on the edge. Keys of the dictionary represent the lower boundaries of 5-dB noise level ranges (45-70 dB) and values exposures as meters. This attribute is not needed if quiet path routing is disabled from the configuration.
gvi	g	float	yes	None	Green view index (GVI) of the edge (0-1). This attribute is not needed if green path routing is disabled from the configuration.

// sample GraphML data
...
<node id="n1905">
  <data key="v_geom">POINT (25497688.29600531 6678110.889221223)</data>
</node>
...
<edge source="n2952" target="n4368">
  <data key="e_ii">1444</data>
  <data key="e_geom">LINESTRING (25497712.53494933 6677892.654694145, 25497714.36057202 6677921.944805327)</data>
  <data key="e_geom_wgs">LINESTRING (24.9587395 60.21378910000001, 24.9587721 60.214052)</data>
  <data key="e_l">29.347</data>
  <data key="e_b_st">0</data>
  <data key="e_b_ab">1</data>
  <data key="e_bsf">1.5</data>
  <data key="e_n">{65: 14.67348, 70: 14.67348}</data>
  <data key="e_uv">(2952, 4368)</data>
  <data key="e_iw">1095</data>
  <data key="e_g">0.23</data>
</edge>
...

Building a custom graph: overview / demo

The scripts for building a custom graph are not generic enough to work for any geographical extent without modifications. Localization is needed for adapting to available data sources, file types, nodata areas, projections etc.

However, it should be possible to construct a graph without noise and greenery data with considerably smaller effort by running just steps 1, 2 and 5. The package-level configurations (config.py files) need to be adjusted to match the local programming environment and data files/directories. Also the scripts may need some minor adjustments, such as setting an appropriate EPSG code for a local projected coordinate reference system.

Running the tests

$ cd src
$ python -m pytest graph_build/tests/otp_graph_import/ -vv
$ python -m pytest graph_build/tests/graph_noise_join/ -vv
$ python -m pytest graph_build/tests/graph_green_view_join/ -vv
$ python -m pytest graph_build/tests/graph_export/ -vv

Prerequisites

• Java Development Kit, preferably version 8 (AKA version 1.8)
• OpenStreetMap (OSM) data file for the area of interest in pbf file format
• Python environment defined in env/conda-env.yml

1. Create initial street network graph from OSM with OpenTripPlanner

1. Place your OSM data file (*.pbf) to directory graph_import/graph_data_in/
2. cd graph_import
3. sh osm_2_otp_2_csv.sh

This should produce two files to the folder graph_import/graph_data_in: edges.csv and nodes.csv.

2. Import & re-build street network graph for Green Paths from CSV files

Demo: otp_graph_import/otp_graph_import.py

3. Join noise data to street network graph (optional)

Demo: graph_noise_join/noise_graph_join.py (uses traffic noise data for HMA)

4. Join greenery data to street network graph (optional)

Demo: graph_green_view_join/graph_green_view_join.py (requires PostGIS, uses land cover and green view index data for HMA)

5. Export graph to GraphML file with only required attributes

Demo: graph_export/main.py

Environmental data for Helsinki Metropolitan Area (HMA)

• SYKE - Traffic noise modelling data from Helsinki urban region
• Traffic noise zones in Helsinki 2017
• OpenStreetMap
• Green View Index (GVI) point data
• Land cover data