[Talk-GB] Modal Filters and Low Traffic Neighbourhoods
Martin - CycleStreets
list-osm-talk-gb at cyclestreets.net
Sat Jun 12 19:53:42 UTC 2021
> I think there are people who have come up with an algorithmic way of
> finding LTNs.
We (CycleStreets) have been working on this over the last few months.
I'm pleased to say we now have a beta map of locations available
specifically for the purpose of OSM mappers looking at it to find cases
where OSM data is missing or incorrect. (Coverage is currently UK &
This is not intended for general consumption at this stage - we are keen to
see OSM data fixes in place so that the analysis picks up missing modal
filters, and errors eliminated, before a more public map release.
Please have a good look round your area, to find cases where there exists a
modal filter on the ground, but we are not picking it up. Please update the
data in OSM :) Changes should be reflected in our analysis a few days
later. I've listed at the end of this message the typical map data errors
I've found while validating the data, which mappers should particularly
look out for.
We know that the analysis is not yet 100% perfect, and some locations we
know are missed for particular reasons. I have noted the current known
shortcomings in the analysis at the end of this message.
MODAL FILTER TYPES
The term 'modal filter' is not an exact term, and there are various
potential interventions that could be covered. We have focussed on those
ordinarily understood to create LTNs, i.e. locations that genuinely create
filtering for cycling and walking, but which are not just general-purpose
walking/cycling links between areas.
1) Bollards in the middle of a highway
This scenario creates a gap through which cycles may pass but not motor
vehicles. (NB We do not include these on service roads, which tend not to
Example - Walnut Tree Way, Cambridge:
2) Gap in the middle of a highway
This scenario is similar but uses a gap. We have set a low threshold to
avoid this picking up cycleways/walkways between streets. We also have a
range of heuristics to avoid finding snickets between parallel roads and
Example - Hulton Street, Salford:
3) Gate in the middle of a highway
Similar to scenario 1 and 2, but a gate is used, with a gap at the side. We
have tried to eliminate cases where there is a gate providing limited
access to a private estate or commercial area, even if the gate hasn't been
marked as access=private, but this is hard to do 100% reliably. Please
ensure that gates are properly marked as private.
Example - Adam and Eve Street, Cambridge:
4) Continuation cycleable
This is where a sideroad doesn't quite meet the main road, but a
continuation allows people to cycle/walk through.
Example - Melville Road, meets Holyhead Road, Coventry:
5) Continuation walkable
Similar to scenario 4, but these are only walkable. (In many cases I
suspect these are actually cycleable as there is no 'No cycling' sign, but
bicycle-yes is missing in OSM.) I believe there are probably a lot of cases
in OSM where the pedestrian link from the sideroad to the main road are not
present. Please do check thoroughly in your area!
Example - Laverna Dale continuation, Dublin:
6) No entry except cycles: contraflow
This is where a 'No entry' sign has an 'Except cycles' plate added,
resulting in motor vehicles not being allowed to enter an area. These are
relatively new in the UK as the DfT only officially allowed such signage
about 10 years ago. We have avoided the marker showing the contraflow being
present at every point a sideroad joins to it - we've tried to get just the
entrance from main roads.
Example - Malcolm Street, Cambridge:
7) No entry except cycles, using a plug
Similar to scenario 6, this is the older arrangement where a physical gap
exists, providing an entry for cycles into an area, allowing them to bypass
a No entry sign to the right. (We cover both the scenario of the drawn-in
cycleway joining to the same node as the junction itself, as well as the
scenario of it drawn connecting to the main road but as a different node
just to the left of the junction.)
Example - Garlic Row, Cambridge:
8) Bus gate
This is where bus lane cameras are used to create a very short section of
bus lane, through which only buses and cycles can legally pass without
being fined. (We have tried to avoid cases where a long general bus lane
alongside other traffic lanes contains a tiny section of Way.)
Example - Silver Street, Cambridge:
TYPICAL ERRORS IN OSM DATA
Typical issues in OSM data I have found while trying to validate the
analysis are as follows - please do have a look round your area and check
for OSM data fixes where you know there is a modal filter (as defined as
one of the above types).
- Missing pedestrian/cycle links from the edge of a local street that
almost joins a main road (see the 'continuation' scenario cases above).
- Gaps that physically include bollards but no bollards are present in OSM
at one/both end(s).
- Gates not marked with access=private.
- Bollards that have not been marked as bicycle=yes.
- Missing contraflows.
- Uses of highway=pedestrian where bicycle=yes is not explicitly stated,
despite it being obvious that cycling is not banned.
- Things where the geometry actually matches a modal filter, but are in
areas which should be marked as access=private (and therefore not really
modal filters insofar as general road users are concerned).
- It would be useful if continuation cases could have an est_width added,
so we can in future eliminate cases that look like continuations but never
actually had direct car links, i.e. genuinely always were just narrow
footways. We already have a rule of thumb on this based on distance, but it
would be useful to have this extra data.
- General connectivity issues where a sideroad has a continuation to a main
road that has a cycleway along it, and the sideroad cycleway only connects
to the main road cycleway and not the road itself also.
SHORTCOMINGS IN THE ANALYSIS
The analysis is quite complex (basically pattern-matching using some quite
complex SQL statements) and we haven't yet zapped every bug.
- We haven't managed to eliminate cases of gates leading to islands of
isolated roads (i.e. private housing estate or commercial estate) where
that island network actually includes a walkway providing a second entry
point. Please ensure such gates are marked as private so these can be
explicitly known to be private entrances.
- Sometimes a bus lane that includes a very short section between the two
sides of a dual-carriageway ends up being shown as a bus gate.
- In the cases where a continuation (local street almost touching main
road) has a walkway, if the main road has a pavement drawn in as a separate
Way, and the continuation walkway joins only to that pavement and not the
main road itself, we are not currently detecting it. This is the problem
that the pavement is not strongly associated in formal connectivity terms
with the main road itself.
- We have a declumping mechanism that aims to de-duplicate cases, so that
for instance a short gap with bollards at either end doesn't end up with
potentially three matches (gap, bollard, bollard). This currently has a
slight side-effect whereby completely different modal filters that happen
to be very close physically have one of the two zapped. We plan to add more
intelligent rules to avoid zapping combinations that logically are very
unlikely to be the same thing. An example is Bene't Street in Cambridge
(cycle contraflow), at the junction with King's Parade (gate barrier):
- We aren't yet dealing with the scenario of gates at the edges of parks,
which need to be eliminated. There are genuine cases of modal filters
within parks which obviously need to be kept.
- There are some edge-cases whereby a continuation actually has multiple
spurs coming off it from the same node, and none get picked up, e.g.
Riverside Cambridge where it joins Abbey Road:
- If a filter has been (pointlessly) chopped into two ways, we may not
detect this. Example at Poplar's Road, Walthamstow:
Martin, ** CycleStreets - For Cyclists, By Cyclists
Developer, CycleStreets ** https://www.cyclestreets.net/
More information about the Talk-GB