Congestion Network 2: Validate Network

[chmnata]

Listed layers that needs to be validated and suggest validation methods.

Network Segments congestion.network_segments

Layer of the congestion network segments

• Check for length, see if there are super short or super long segments and if they are valid

• Check for link_dir that makes up a segment, are there any cases of duplicated link_dir?

• Is the length correct? Summing up link_dir's length to double check congestion.network_links_22_2 (the lookup table)

• Is the highway label correct? Are the segments labeled as highway actually highways? (can check with centreline)

• Are there any segments with no direction tagged?

Network Nodes congestion.network_nodes

Layer contains nodes that make up each network segment

• If all the nodes are used in the network segment table

• Are there duplicate nodes?

Centreline int_id and HERE node matching network_int_px_22_2

• If all the nodes have a matching for either px or int_id

Centreline segments congestion.centreline_segments

Layer contains sets of geo_ids that represents each segments.

• Compare the length of the segments and the length of the centreline, there is a lookup table created in congestion.segment_centreline_lookup that has segment set length and centreline set length for comparison

• If segment_ids were used more than once incorrectly

Centreline segments congestion.centreline_segments_name

Layer contains sets of geo_ids that represents each segments + Street name

• Are the names correct?? There are arrays of street_name, if its more than 2 its kinda weird and should be investigated

• any null name ?

[radumas]

Might add identifying segments with low/no data and doing a little investigation per https://github.com/Toronto-Big-Data-Innovation-Team/rapidto/pull/38#issuecomment-1284227182

[radumas]

I think two of these tables are names wrong:

• congestion.centreline_segments is congestion.segment_centreline
• congestion.centreline_segments_name is congestion.segment_centreline_name

[radumas]

There are concerns that some meaningful segments of streets may be missing due to road classification.

• Investigate the proportion of each street name that is part of the network.

[Nate-Wessel]

Issues found from manual validation of congestion.network_segments so far:

• Bridges don't seem to be accounted for; segments are broken where they cross other segments they don't actually intersect with due to vertical separation. E.g. Prince Edward Viaduct over Bayview Ave, DVP - lots of other cases too though.

• id 387: Vaughan is two-way here, but there's no paired edge going the other way
  

• ids: 3570, 3571, 3572, 164, 165, 342: Intersection of Park Lawn and Lake Shore West needs to be de-duped. There are two SW-bound edges, only one of which connects to Park Lawn.
  

• Missing connection at Dundas / Bloor / Kipling
  

• ids 166, 3569: stub segment
  

• ids: 1692, 4885, 4888: Wellington Street stub with erroneous 2-way:
  

• Missing section of Morningside at Finch:
  

[Nate-Wessel]

Should I be looking at congestion.segments_v6 instead of congestion.network_segments? segments_v6 seems to correct some of the problems I described above.

[Nate-Wessel]

not all highways are marked highway

[ShahrzadBorjian2]

I found a few issues with congestion.segment_centreline_lookup table.
The table contains incorrect start_vid or/and end_vid for the segments that intersection names of their end points are missing (NULL) in gis.centreline_intersection table.
It seems int_ids of the closets intersections with available intersections name have been assigned to these segments.

Please see the following example from the CafeTO project for more clarifications and details.
I used the following sql to get intersection names of the segments within the study area (cafe_to.“TEST_st_clair_west" ) to create lookup table with intersection names.
• Joined cafe_to.“TEST_st_clair_west" table and congestion.segment_centreline_lookup using segment_id to get start_int and end_int ids
• Then joined gis.centreline_intersection using int_ids to get intersection names

As it can be seen in the output table, some segments such 1172 and 1173 have incorrect vids. Their start_vid and or end_vid represent segment sets rather than individual segment.

with start_node  as 
	(select * 
	from cafe_to."TEST_st_clair_west"
	left join congestion.segment_centreline_lookup b 
		using (segment_id) 
	left join gis.centreline_intersection d 
		on (b.start_int= d.int_id)
	)
select distinct 
	segment_id, 
	start_vid, 
	end_vid,
	start_int,
	end_int, 
	start_node.intersec5 as start_int,
	e.intersec5 as end_int,
	segment_list,
	dir,
	segment_geom
from start_node  
left join
gis.centreline_intersection e  
	on (end_int=e.int_id),
lateral congestion.get_segments_btwn_nodes(start_vid::int,end_vid::int) 
	results(start_point, end_point,segment_list,length,segment_geom)
order by segment_list

[ShahrzadBorjian2]

I noticed some missing segments in the congestion network while I was creating study areas for the CafeTO project.
Segments are broken/missing when a road class changes resulting in disconnected corridors. Please see the following example in the Downtown East study area:

Study Area:

Missing Segments:
there is no segment available in congestion.network_segments along Jones and Greenwood between Dundas and Queen.