OSM Park Fingerprinting
Linus Dietz
2023-02-15
General Overview
There are 3451 parks in London when querying for
leisure:park. 1861 of the parks have a name. In the next
step, we collected all map objects (nodes) within these parks.
## name node_count
## 1: Dulwich Park 2345
## 2: Peckham Rye Park 2129
## 3: Hampton Court Park 1594
## 4: Burgess Park 1348
## 5: Russia Dock Woodland 1316
## 6: Queen Elizabeth Olympic Park 1217
## 7: Hyde Park 1183
## 8: Richmond Park 985
## 9: Victoria Park 909
## 10: Southwark Park 869
## 11: Bushy Park 797
## 12: Kensington Gardens 749
## 13: Burgess Park West 731
## 14: Peckham Rye Common 600
## 15: Queen Square Gardens 525NB: It is also possible to query for rivers/lakes etc., but there are typically no nodes within bodies of water.
Heatmaps
These heatmap shows the density of map objects within the parks. The interactive heatmap is also available: https://lynyus.org/_maps/parks_nodes.html
London
Regent’s Park
Queen Mary’s Garden
Relevant Categories
This is a bit more complicated, since it turns out that categories are generally tags in the form of a {key:value pair}. OSM editors should orient their submission according to the general guidelines, but in practice these are not perfectly enforced and these tags can be freely edited. Although obscure tags are not displayed on the map, we still need to deal with them when using the API.
Currently, I have focused on the keys of the tags and made a decision
whether they seemed to be relevant for me or not. The full list can be
viewed (and edited) in keys_relevancy.csv.
In total, I discovered 613 distinct keys in the parks of London, of which I found 68 to be relevant.
To Do: also filter for tag values
## Key relevancy
## 1: amenity 1
## 2: archaeological_site 1
## 3: artwork 1
## 4: artwork_subject 1
## 5: artwork_type 1
## 6: athletics 1
## 7: attraction 1
## 8: bar 1
## 9: bench 1
## 10: bicycle_parking 1
## 11: bicycle_parking:stands 1
## 12: bicycle_rental 1
## 13: brewery 1
## 14: building 1
## 15: bunker_type 1
## 16: cargo_bike 1
## 17: cemetery 1
## 18: community_centre 1
## 19: craft 1
## 20: fast_food 1
## 21: fireplace 1
## 22: fitness_station 1
## 23: fountain 1
## 24: golf 1
## 25: heritage 1
## 26: hiking 1
## 27: historic 1
## 28: landmark 1
## 29: landuse 1
## 30: leisure 1
## 31: man_made 1
## 32: memorial 1
## 33: military 1
## 34: monument 1
## 35: museum 1
## 36: natural 1
## 37: office 1
## 38: outdoor 1
## 39: outdoor_seating 1
## 40: park_ride 1
## 41: parking 1
## 42: pedalboat_rental 1
## 43: place 1
## 44: playground 1
## 45: power 1
## 46: public_transport 1
## 47: pump 1
## 48: railway 1
## 49: recycling 1
## 50: recycling_type 1
## 51: religion 1
## 52: ruins 1
## 53: seating 1
## 54: shelter 1
## 55: shop 1
## 56: site_type 1
## 57: species 1
## 58: sport 1
## 59: station 1
## 60: street_lamp 1
## 61: subway 1
## 62: toilets 1
## 63: tourism 1
## 64: utility 1
## 65: waste_basket 1
## 66: water 1
## 67: waterway 1
## 68: zoo 1
## Key relevancy## Key relevancy
## 1: FIXME 0
## 2: HE_ref 0
## 3: abutters 0
## 4: access 0
## 5: access:conditional 0
## ---
## 541: wikipedia 0
## 542: wikipedia:en 0
## 543: wpt_description 0
## 544: year 0
## 545: year_of_construction 0Park Fingerprints
My idea would be to characterize a park by the nodes within its boundaries. We could come up with scoring functions where each node contributes with a certain weight to scores in different categories:
- A bench scores 1 point for amenities
- A café scores 3 point for amenities
- A restaurant scores 5 point for amenities
- A tree scores 1 point for natural
- A fountain scores 5 points for art
- etc.
The weights of the different items could also be adjusted using blackbox learning to approximate a concept that we want to emulate (I did something similar here: https://www.frontiersin.org/articles/10.3389/fdata.2022.829939/full).
This is an excerpt of some categories, as mentioned there are currently 68 relevant tags.
## name natural leisure tourism artwork_type playground
## 1: The Regent's Park 129 9 28 5 2
## 2: Queen Elizabeth Olympic Park 668 14 98 10 5
## 3: Hampstead Heath 46 0 16 2 0
## 4: Kensington Gardens 464 0 30 7 0
## 5: Battersea Park 134 7 14 2 0
## 6: Victoria Park 562 45 27 2 11
## memorial historic amenity sport landmark bicycle_parking
## 1: 0 0 335 0 0 11
## 2: 3 5 312 39 0 45
## 3: 6 7 249 0 0 3
## 4: 2 4 239 0 1 7
## 5: 3 4 234 0 0 16
## 6: 2 4 229 34 0 13Improvements
- Compute area of parks to compute density measures (requires tessellation)
- Compute land use of parks [grass, wood, water, flowerbed, garden,
sports pitch, playground, etc.]
Access to parks
3.25% the parks do have an access tag indicating whether
they are accessible. 49.11% are generally accessible, 49.11% are
private.
parks_access[, .N ,by = access]## access N
## 1: private 53
## 2: yes 44
## 3: 3338
## 4: permissive 11
## 5: no 2
## 6: unknown 1
## 7: agricultural 14.75% have a tag indicating the opening hours.
Discusssion
- What would be a compelling research question around the theme of Park Fingerprinting?
- Is there any use case this could be good for (e.g., extensions to existing projects)?
- What could we tell about the urban greens in different cities globally?
- Would it be interesting to pair this data with satellite data? If so, what could we show?