OpenStreetMap

SK53 has commented on the following diary entries

Post When Comment
New road style for the Default map style - the first version 7 months ago

@Mateusz:

The OS map example is excellent, because it also shows you things which are more or less impossible to read even on the 1:25k cartography of the OSGB. Notably that there are footways between some of the roads in the top right-hand corner. This is a personal bugbear of mine with the Explorer maps: it is more or less impossible to use them to navigate in urban areas which are rich in paths. For instance, about 5 years ago, I was in Corby and needed to walk from the centre to a hotel on the outskirts. Much of Corby was blank on OSM so I used my Explorer map: the safe route I took was around 1.5 km longer than the shortest route, because I did not want to risk finding myself in a dead end. Currently OSM standard rendering is much more satisfactory for this purpose than OSGB: even if OSM was not designed at the outset for such a purpose.

In practical terms this is because OSGB shows hedges & fences in urban areas which tend to create too much clutter for things like paths to be readable. It does make for an attractive product, but they certainly have scope for improving this: for instance subtle difference in style for urban/rural areas (which is something we should consider for OSM).

New road style for the Default map style - the first version 7 months ago

Hmm, markdown for images doesn't seem to have worked. Here are links:

Red Motorways on OSGB 7th series: https://flic.kr/p/vuH9ob

Blue Motorways on AA Road maps: https://flic.kr/p/gNjieo

New road style for the Default map style - the first version 7 months ago

With respect to the motorways are blue issue. It might be interesting to look at the history of this colour selection by the Ordnance Survey.

Historically the 1 inch series used red for primary roads ('A' roads) whether they were designated trunk or not: the difference was shown by a "T" in brackets after the road number.

OSGB 7th Series Motorway rendering

Here's the same area on current OSGB Landranger via Bing Maps.

When Landranger 1:50k maps were introduced in 1974 one of the major changes in appearance was that motorways became blue. I am sure there were three factors behind this: the maps had become much less useful for both motorists & pedestrians as the motorway system developed; blue was also the colour of motorway signs and thirdly, for cost reasons, they needed to use one of the existing colours (IIRC 7th series used a 7 colour print process). I am sure they considered any implications of confusion with rivers at the time.

I can't remember much fuss about this change: there was a considerable brouhaha about administrative borders and path symbols being too similar (which led to a change quite quickly).

Trunk roads became green much later (and trunk & primary are still shown with similar styles on the more walker-oriented 1:25k maps), but I would think the motivation was similar: to make the maps more useful/appealing to motorists.

I'm not absolutely sure, but I think blue for motorways may have first appeared on AA maps. Here's one example from the mid-1970s:

AA map with blue motorways

shops as closed-Way building outlines, but also as Nodes in the center? 8 months ago

As someone who has been working to create a global POI dataset of retail outlets I can confidently state that this is fairly straightforward if one uses osmconvert (or for more limited data overpass with the centroid feature).

Processing the data to remove duplicate node/polygon elements turns the simple code into something far more complex. Interestingly following MarkusHD's approach means that things like address data can be incorporated trivially in the data set. So I'm happy with duplication of addresses (straightforward as we, like many others, treat addresses as attributes of elements, rather than elements themselves), but vehemently object to duplicating elements themselves.

"Duck Lanes" on OSM? 9 months ago

Please dont map animal tracks in general & certainly not as paths. First in some places there are many many animal tracks (mountainsides with sheep come to mind). Secondly, animal tracks tend to change depending on circumstances. Thirdly, mapping lots of traces which are not necessarily usable by people will just make OSM data hard to use.

Thoughts on the import of address nodes in JOSM using MassGIS L3 Parcel data to aid in adding addresses to buildings in MA 9 months ago

Virtually all the MassGIS imports have suffered from major and significant problems. Firstly the import of streets did not join ways at boundaries. As the import was on city/town boundaries rather than county elsewhere making MA routable was extremely tedious.

Current parcel boundaries do not seem to be massively accurate. For instance this playground in Yarmouth Port is not contained within the two forks of Playground Lane and Old Church Road. It's a while since I last visited, but Google StreetView & Bing aerial imagery confirm that the MassGIS boundary shown on OSM is inaccurate. A little to the East beyond the end of Damaris Drive there are multiple overlapping polygons for wetland & conservation areas: very difficult to work out what they are doing. I've tidied some of these up using aerial imagery/memory in the vicinity of Bass Hole/Grey's Beach, but my impression is many of these polygons are more general than we expect with OSM.

Another area with problems in the past was the Hyannis Transportation Centre. Most buildings had been imported, but as they were imported some considerable time after the roads; the new road network was not in place. I added this (and fortunately Lars Ahlzen has done more) when trying to put the P&B bus route in.

The significant problem with these imports is that everything looks complete, but in reality it requires considerable revision, checking against current status, and judicious removal of useless data (see the large number of tags on the playground above). Very large numbers of highway=residential are nothing of the sort, but most likely driveways to properties (certainly along the Kings Highway in Barnstable, Yarmouth Port, & Dennis places I'm reasonably familiar with).

A side effect of this is if one looks at POI density in the Upper Cape, it is woeful. Only 28 restaurants mapped in the Barnstable admin division, which includes multiple popular holiday resorts including Hyannisport. (Only 1 in Rockport, 4 in Gloucester, 1 in Plymouth, 8 in Harwich, and NONE on Nantucket, etc). It doesn't help that GNIS nodes are often inaccurately placed (such as Friends Meeting House on Nantucket which is several streets away from its actual location), and close to its likely location MASS GIS buildings sit on top of MASS GIS roads.

All in all these are reasons why I would be extremely cautious about further mass imports of data for Massachusetts. I cant be certain, but my basic feeling is that at least POIs would be better represented in the absence of these imports. Adding housenumbers would give further weight to the appearance of a complete and detailed map: but in practice in places I know, OSM is much less useful than it could be.

Alternative processes would be to put your data set up somewhere so that local users could use this as the basis for mapping (e.g., as a tiled layer or similar), or simply make it available in such a way that it can be integrated with OSM data at the data consumption stage, so that applications can use it (in practice geolocated address nodes may be all that is needed)

Vikumberg, Fünfenberg, Tabor nad Botačem, Draški Tabor, Tabor di Draga, Castello di Moccó, Lorencan, Sv. Lorenc 9 months ago

Great bit of detective work!

Melbourne tram stop project - first post 10 months ago

You might like to take note of Mapillary. Open Data from VicRoads is being progressively loaded onto Mapillary: this may really help add details without committing to visit every tram stop.

I cant find an example with trams because many of these sequences have not yet been fully processed by Mapillary and dont yet show up as a blue line on the map.

#1009 - Nepal Earthquake, 2015, Gorkha 10 months ago

Some of the aerial imagery, particularly the grey scale MapBox imagery is quite difficult to interpret and in particular I have noticed that ridge lines and valley bottoms are often initially difficult to pick out.

The two ways to check are a) use the OpenCycleMap layer which shows contours and b) zoom out until you have Landsat satellite data. With the latter the overall topography becomes much clearer. (I may put a bit on the wiki in the next day or so now that you've raised it).

In areas where this is particularly difficult to interpret I actually crudely map ridge lines (natural=ridge) as well as waterways. This not only provides data, but helps to keep the context of the imagery clearer. I discuss this a little on my blog about mapping Lesotho.

How many amenities are open for a given time? 10 months ago

Nice thoughts.

I think the graphs are confusing because of the variable axes. In practice the vast majority of entries are parsable. I'd suggest making the % axis from 0-100, and perhaps adding another line showing total parsable entries using the right axis.

What I dont get a feel of is how many elements might have opening hours tags on them vs those that dont: for instance most shop=* might be expected to be valid elements for opening_hours.

Another thing which one might be able to pick up, in Germany at least, is the local Ruhetag, just from which day typically has different opening hours in an area.

Potlatch and contour lines 10 months ago

I imagine you used OpenCycleMap as a background layer. This no longer appears in the default layers. I dont know if this was deliberate, or accidental. I imagine you can still create this as a layer for Potlatch2.

The town with the most toilets on the world! 10 months ago

I doubt if the things tagged as toilets in Tandale are private toilets, some are probably Open Defecation Areas, others might be communal toilets. Most of the 50k+ inhabitants will not have access to any kind of private toilet facility. An overpass query counts around 550 POIs, so roughly 100 people:1 toilet.

Tandale has been mapped by local people with guidance from experienced OpenStreetMappers.under a Development Seed programme.

I suggest reading some of the following:

And a youtube video of Mark Iliffe and Gary Gale made in Tandale.

Hamlets in US cities 11 months ago

@lxbarth This sounds a) a great typology and b) good sense; but it raises one of my long standing tagging concerns. Just what is the best way to tag trailer/mobile home parks over-and-above just using landuse=residential?

Hamlets in US cities 11 months ago

@lxbarth This sounds a) a great typology and b) good sense; but it raises one of my long standing tagging concerns. Just what is the best way to tag trailer/mobile home parks over-and-above just using landuse=residential?

Hamlets in US cities 11 months ago

Nice post, this has been a personal bugbear for ages. I'd love to see a concerted effort to fix this.

The obvious alternatives are place=suburb & place=neighbourhood (possibly place=locality, but like hamlet I expect that in rural areas). A third was is just to place a landuse polygon of the appropriate type and add the name tag

Many of these are industrial/commericial areas with a low population (hence why they got the hamlet tag), but in some cases it's just that the naming of areas in the city is at a fine degree of granularity.

It's not because you have accurate data that you have to upload all of them in OSM 12 months ago

And we have really accurate position of the pins on the golf course, which will of course get moved every few days to keep green wear constant.

Who are the Missing Mappers? Part II about 1 year ago

Fundamentally I agree with this, but I also think that practical surveying experience greatly assists in interpreting images even in unfamiliar places.

I know that my initial mapping in Port-de-Prince was very much informed by how much I had learnt from adding stuff remotely and then discovering that in reality things were rather different. There is always a danger of over-interpreting remote imagery, and exposure to this issue by actual survey experience really helps.

In the short time allotted for a Missing Maps session practical surveying is obviously impossible, but I think there is scope for using things like Mapillary sequences etc., to place remote mapping in a context together with on-the-ground surveys.

Years ago nickw and I spent half a day presenting OSM at a MapAction training weekend. I remember being surprised by how many of those present had edited OSM in their local area. Perhaps this was besides being GIS pros, many wanted a feel for how data was created. After all, they use it in earnest during humanitarian crises.

Casas de sucos about 1 year ago

I dont think it's quite right, but will do for a start, providing you add some other tag which will enable you to find juice bars. In my case I've added cuisine=milkshake as such a start.

Casas de sucos about 1 year ago

There is a (small ?) chain called Shakeaway in the UK and elsewhere selling Milkshakes. These have been tagged in several different ways: amenity=cafe, amenity=fastfood, shop=yes, shop=beverages. So we could do with an answer too. There are close analogies to ice cream places (not exactly food, can consume in or take away).

Learn-a-tag: leisure=fitness_station about 1 year ago

Certainly noticed several in Parque Centenario in Buenos Aires. There is also a very little used tag covering several of these leisure=outdoor_gym. In Switzerland and nearby there are specific trails with fitness stations located along the trail. There has been an ongoing project to map them for several years: ParscoursVita.