Field verified OSM data: most useful tags to collect

General talk
1

Why does field-verification matter?

  • A quick pre-amble to address: why does this matter?
  • Commonly OSM data is mapped remotely, using aerial imagery of some kind.
  • ‘Mapathons’ may be held - using satellite or drone imagery to create mostly road and building footprints - to rapidly populate the OSM map in historically poorly mapped areas.
  • Field mapping is required as a ‘ground truth’. Although the feature may exist via imagery, key questions are (1) has the feature changed since the imagery was last updated? (2) does the mapping in OSM match the reality on the ground?
  • There is no better way to confirm that a feature exists in reality and is documented correctly, than to actually see if first hand in the real world.

Field-verified OSM data via FieldTM

  • I work for HOTOSM, and we * won a microgrant via the OSMF for some * updates to FieldTM.
  • The * Field Tasking Manager (FieldTM) is a tool to help coordinate open mapping campaigns in the field: when a group of people either (1) verify existing data in OSM in real life (2) add missing data to OSM, present in the field but not on the map. See the * FAQ for more details on FieldTM vs StreetComplete vs Everydoor etc (summary: it’s not a competitor, but a complementary technology).
  • Currently we have a * collection of field mapping surveys(in XLSForm spreadsheet format) that map 1:1 with OSM tags, for post-processing after field mapping and upload of new data to OSM.
  • The current forms are more ‘humanitarian’, rather than general OSM community mapping focused.
  • Note that we are also working alongside the GFZ Helmholtz Centre for Geosciences (cc @pomodoren) to create forms for Exposure Mapping (assessing threat from natural /environmental hazards) and getting a portion of data back into OSM too, for which there will be a talk at FOSS4G Hiroshima if anyone is interested.

Asking the community: what do you want field verified?

  • As mentioned above, the current forms are more humanitarian in focus, with our initial goals focusing on our * priority countries.
  • As part of the microgrant, I am asking here for feedback on what the OSM community at large would like to see included in the default forms available in FieldTM.
  • The perfect scenario here would that local OSM chapters and mapping communities globally adopt FieldTM to collaboratively map areas of their communities: a nice way to socialise and keep active, but primarily to update the OSM map.

Key questions:

  1. What are currently the biggest pain points within OSM for data that is commonly mapped remotely, but not actually verified to exist in the field?
  2. What types of features frequently change on the map and need to continually be updated by field observation (by OSM communities and local mapping groups)?
  3. Are there any essential OSM tags missing from the already created (linked) survey form categories?
  4. Are there any feature categories that are omitted entirely, that the community would love to see included as a pre-canned ‘OSM xxx Features’ survey?

I really welcome input from anyone here, and hope to stimulate open discussion within the community!

2 Likes
2

up-to-dateness

Shops, opening hours, access rights, existence of objects,

Sorry, that’s far too much to go into; it would be beyond the scope of this discussion. But for me, StreetComplete – or SCEE – is the perfect tool for on-the-ground / field data collection. It already covers everything I consider worth mapping when I’m out and about.
Road classification, surface, smoothness, access restrictions, opening hours, business types, and POI existence are my most contributed topics when surveying. But also roadworks, road closures, house numbers, obstacles, gates, entrances, road infrastructure…

1 Like
3

What do I want field verified? All of it.

So what might really matter to me will depend on my particular special interest which comes and goes.

It might be bus stops, AEDs, public toilets or pubs with great beer. Some of those might be of interest at the same time. Hopefully not the AED.

I would guess that there is unlikely to be an easy answer as we all have different needs.

But I’ll take a punt at field mapping and verification of all dropped kerbs and driveways etc to enable correct sidewalk mapping from aerial imagery sources.

1 Like
4

Have you looked at StreetComplete as a base set of information that’s easy to collect with in person surveys?

If you have not already checked the existing implementation, I don’t see much point in redoing tag collection with a forum survey.

Specifying the type of a kerb=yes is extremely easy to do in StreetComplete.

EDIT: I now see that I skimmed too hard and it is mentioned, and you’re talking about how to assign geographic areas for in person survey.

1 Like
5

Nederland probably has a luxury position - our datasources to copy from include buildings and addresses data, traffic signs data, road data, waterways and waterbody data, landuse data, public transport data, properly aligned satellite imagery and aerial imagery, elevation and height data, and more; all of which are official and fairly recent, say one year at most. Even trees, benches, waste baskets and other street furniture, manholes, fountains, guard rails, culverts, playgrounds, walls and hedges. These sources are, on average, much, much more accurate and recent than most survey data wil ever be, simply because in the field we cannot measure things as well as the official bodies maintaining and providing the data. The sources do contain errors, but estimated 98,237% is correct. Survey: most of the osm-elements entered from survey can usually be easily recognised by the lack of precision. I recognise some mappers who enter groups of trees during survey, by the manga hairstyle outlines of these “forests”, placed somewhere in a polder, next to or overlapping similarly styled water bodies.

So what’s missing is mostly extra details which we can’t see from the air: materials, colours, construction of things, names of shops and businesses in the buildings, opening hours; and minor footways and cycleways in privately owned or permissive areas, where we have no government data.

1 Like
6

All very valuable context - thanks for the inputs so far!

So the common theme is that field mapping should be done for features you can’t reliably pick out from imagery, or gov datasets. The gov datasets part is only really applicable to high income economies, a fraction of the total global population, but it’s still a very good consideration for field mapping in these contexts :smiley:

So far

  • Hidden details in transport: sidewalk curb types, driveways and sidewalk connections, small footpaths / cycleways.
  • Useful amenities that are missed on the map: public toilets, bus stops, AEDs.
  • Regularly changing building details: shop name, business type, opening hours,
  • Descriptive attributes: colours, surface types.

Regarding StreetComplete, I’m definitely on the bandwagon - it’s a fantastic tool!

FieldTM is a complementary tool, built for coordinating a mapping campaign (say you have 100 people collaboratively mapping at the same time in a city - you will have effort duplication just using StreetComplete alone).

There is a section in the FAQ about this =) I would love if we could find an easy integration surface for StreetComplete, to allow usage as a ‘downstream’ mapping tool during the coordination.

Edit: I did some research on what a possible integration with StreetComplete could look like. It’s not very promising right now, but we could open up the discussion via the repo. EveryDoor is also a possible candidate.

7

Avoid it by allocating people to specific areas. Use something like Field Papers https://fieldpapers.org/, give each person their area/s, then let them fill details in using SC.

Have been on field mapping days using this & it was great, even to the extent of people walking down opposite sides of the same street, waving at each other as they passed! :slightly_smiling_face:

2 Likes
8

You are already starting on the wrong foot here, that is neither historically true, not true for any country that has most of its infrastructure mapped. It -might- be true for a very narrow segment of data, that is buildings.

It -is- true that it is what HOT has massively promoted over the years, particularly with its buildings fixation (see above). So it is good to see you coming around.

There is a, very, limited pool of regular contributors (and fwiw edits), it is down to a percent or so a zero sum game. So yes, it is very much a competitor.

OSM shouldn’t be about you telling other people what to map, unluckily doing that has a very vocal fanboi base. But despite not going to be popular, my answer would still be “nothing”.

4 Likes
9

Two very useful perspectives there - thanks both!

Avoid it by allocating people to specific areas. Use something like Field Papers https://fieldpapers.org/, give each person their area/s, then let them fill details in using SC.

Field papers is a really nice tool: I like and support it!

That said, I personally think it’s more aligned with people that understand maps well & are able to coordinate with a hybrid of paper maps and a digitial tool like SC.

You might think this goes without saying (it’s the OSM community after all!), but you would be surprised by the level of mapping and tech literacy of many potential mappers. People may have very valuable info to contribute, but aren’t very geospatially minded - the simplest fully-integrated solution will generally win out.

A bit more context on where tools like StreetComplete won’t work well:

  1. The geometry is not mapped in OSM.
  2. The geometry isn’t correct in OSM, and the user wishes to correct it.
  3. The geometry is mapped in OSM and doesn’t actually exist.

Comprehensive field verification (the kind that could be done in a structured mapping campaign) should ideally involve both geometry correction and tag addition.

I completely agree with you here! That said, the regions I am primarily focused on mapping do not have ‘most of their infrastructure mapped’. Context is key here: mapping in South Sudan is going to be somewhat different to mapping in Austria.

I have no have all of the context around issues related to ‘HOT’s building fixation’, so apologies for my ignorance here :folded_hands: But in my mind, mapping buildings when no (or very little) map data already exists, seems like a worthwhile exercise to me!

I guess the point I was trying to make is that FieldTM does no actual field mapping. It defers field mapping to downstream tools (QField and ODK in this case, but potentially EveryDoor and StreetComplete are options too). It’s very much analogous to the solution posted by @Fizzie-DWG above, where FieldTM would be more of a competitor to FieldMaps instead (as it simply subdivides and creates projects).

While I also completely agree with you there, I would say there is some utility in having guidelines on what may be most useful to map. Perhaps I’m asking in the wrong place, as the majority of mapping I work on is related to utility: disaster response access routes, electricity infrastructure, which buildings are are at risk from collapse during an earthquake, etc.

Again, this comes down to which community we are asking. OSM France may say “we don’t care what you map, as data quality is already great and users should choose to map whatever they want”. While OSM Sierra Leone may say “we really need to better understand the population and infrastructure in informal settlements around Freetown, so we can better assess their needs and plan the city. A good start would be mapping all the electric infrastructure” (random example).

That all said, I really value the input and opinions of anyone in this thread!

The suggestions most definitely do not need to be humanitarian (or global development) in focus. The whole idea of the exercise was to gather feedback for the OSMF microgrant, to help steer the direction of a field mapping tool that can be used by OSM communities (if they wish).

If in the end people decide they don’t like the surveys available, how the tool works, etc - then I’m not against recommending another project entirely (as long as it can handle the needs / requirements listed above). But let’s not fixate too much on the tools, but instead the field data we wish to collect. In the end, I think we all have the same end goal in mind: making OSM the best quality map it can be :smiley: