User:SK53

From OpenStreetMap Wiki
Jump to navigation Jump to search
No Wikifiddling SK53
ignores tag voting procedures.

Contacts

I blog mainly on OSM related themes at Maps Matter. I tweet as @SK53onOSM, and can be contacted regarding OSM related stuff on Google Mail as SK53 dot osm.

Gear

Map using a Garmin etrex Summit HC, usually with recent OSM maps from User:Computerteddy, Canon camera, Olympus dicataphone, pencil & paper. Most mapping is on foot, so I have a bias to trying to capture the stuff which is non-obvious on even high quality topo maps such as OSGB or SwissTopo.

Useful Garmin Stuff

Many OSM outputs for Garmin are designed for devices with an external memory card. However, my Garmin is limited to about 22 Mbytes of map storage. This is perfectly adequate, allowing complete coverage of places the size of Wales, but does mean that I cant use many pre-built products. Apart from making a plea for developers of OSM Map on Garmin]], I've recently been following this piece of advice [1].

Areas where I map

Ongoing mapping

  • Nottingham
  • Nottinghamshire
  • Lincolnshire Wolds. A certain frustration with trying to match NPE features to the actual Nottinghamshire landscape led me to look for somewhere with less change over the past 50-60 years. The Lincolnshire Wolds were ideal: more or less unmapped, the major new features being WW II airfields (usually easy to pick up on Yahoo, even when long defunct). My aim was to map from NPE: the more obvious roads (presumably tertiary on the ground), streams, rivers and other waterways, parish boundaries, paths. Woodland has been mapped where a toggle between Yahoo and NPE shows similar extents.
  • NPE Parish Boundaries in the East Midlands. East Midlands means former Derbyshire, Leicestershire, Lincolnshire, Nottinghamshire and Rutland. For completeness this should include Northamptonshire. These are usually shown as a pale grey line, although on later maps this changes to a series of closely spaced dots. With the obvious caveat that these may have changed subsequently, these are a very useful starting point for all kinds of administrative boundaries. As they often follow other landscape features (roads, woodland edges, streams etc.) they are also amenable to later refinement when GPS or survey data are available. There is also a lot of historical data relating to original parish boundaries (e.g., Victoria County histories), older county Floras etc., so they are of intrinsic value for mapping. I have replaced a substantial number from OS OpenData BoundaryLine as of June 2010. SK53 18:06, 21 July 2010 (UTC)
  • NPE Waterways. Following the example of Steve Chilton I've been putting waterways sourced from NPE in various hilly areas (notably Yorkshire Dales and Howgill Fells).
    • Currently paused just S of the A66. SK53 12:55, 17 June 2009 (UTC)
    • Mike Collinson has been very active around here updating data and adding lots more features from NPE. So for now I'm holding off until he's passed through! 17:44, 21 July 2009 (UTC)
    • Recently did Kintyre from the OS 7th series. SK53 20:13, 2 November 2009 (UTC)
  • Serre Chevalier ski area and Briancon. Visited here in 2008 so have a reasonable number of photos and have been mapping using cadastral data, landsat etc.
  • Maidenhead. An area with pretty thorough coverage, both in mapping and GPS tracks. I have been endeavouring to validate what exists and fill gaps (mainly POI, footways, short cul-de-sacs). The advent of Flickr supporting OSM tags has led me to post a number of geolocated photos used in this mapping on Flickr. Dair Grant's blog on comparing Edinburgh OSM mapping to Google was the inspiration for this approach (although I'm nothing like as systematic as he was). I am gradually working through these adding appropriate OSM ids on Flickr. A nice way to see them is on Andrzej Zaborowski's OSM/wikipedia viewer.
  • Unterengadin. Mapping woodland/forest from Yahoo from Martina to Zernez, in the Swiss National Park (Val Fuorn), and Val Mustair.
  • Extremadura. This is a huge sparsely populated area with limited mapping activity. I've entered a fair number of waterways, tracks and residential landuse from Yahoo satellite data. Recently I discovered the area around Merida has Yahoo aerial imagery and I have mapped most streets in Merida, and about 5 villages covered by the aerial photos. More can be done, but it really needs on the ground attention. I revisited in February 2010 and was able to add a fair number of tracks, mapping in Caceres, Alcantara,Coria, Guadalupe and Trujillo with some places in between. Irritatingly missed getting traces in Brosas and Montanchez through forgetting to switch trace back on. Xapi download. SK53 18:06, 21 July 2010 (UTC)

One-off efforts

These are generally mapping activities based on a visit, and a set of GPS traces, or community inspired activities based on an event or availability of data. Usually the effort is bounded by availability of data, but for large scale activities (TIGER etc) more likely to be interest and enthusiasm. Often I take the opportunity to learn more about the OSM toolset or to use data which is not available in my regular areas.

  • Monterosa Ski region. Basically the communes of Ayas, and Gressoney la Trinite in the Val d'Aosta, and Alagna in Piemonte. Based on a weeks worth of tracks from March 2009. Recently (June 2010) have been using aerial images to add to this data.
  • Ürümqi [ug:ئۈرۈمچى] [zh:乌鲁木齐] (map). Prompted by a diary entry during the recent unrest. First time I've tried mapping non-natural features in an area I have not visited. There is decent coverage by Yahoo. Very aware that one's cultural background affects interpretation of aerial photos, so likely to be inaccurate in places. Also the Yahoo photos are probably 2-3 years old and thus some of the mapping will be completely obsolete, as in any city in China. One universal problem seems to be how to interpret very broad streets. In many places in the world cities tend to be mapped with a large number of trunk, primary and secondary roads, which tends to lead to unpleasant looking maps. In Ürümqi most large roads have very obvious crossing places, particularly noticeable at junctions. This suggests that many of these streets are not, or only partially divided, and probably only deserve secondary or tertiary status. In the main after a splurge of drawing buildings which I assumed to be blocks of apartments I've been more interested in trying to define areas of landuse. As there is not a single built-up area, rather development happens piecemeal, this can be involved. In some ways it is easier to identify the surrounding agricultural land, recent housing developments and quarries. Industrial and commercial areas are much harder to be sure about: although chimney shadows allows the identification of heavy industrial plant. Minor roads are also hard to pick out: partly because the photos were taken at an angle and higher buildings often obscure details which allow roads to be recognised (parked vehicles, traffic). All-in-all an interesting experience.
  • TIGER Fixup. Managed to get involved in this without realising that it is pretty humdrum and nothing like as satisfying as tracing wholly new areas from Yahoo. The rendered map doesn't look that different after a lot of work. Most of the difference is in routability and at the moment this only shows up on the Cloudmade website a few days after the weekly planet dump. I've mainly been interested in sorting out areas which I know, or are connected to other interests. So I've done work around Cape Cod and elsewhere in the Boston area, in Colorado, Dover, Delaware and along I-25. What is interesting is how one can use databases (Postgres) to look at interconnectivity of OSM data. So I've spent quite a lot of time trying to set up lots of OSM tools in a Windows environment. I've also been trying to implement old algorithms of mine for graph traversal in SQL. I'm interested in using them to look at waterways as they allow traversal of many (thousands or greater) of discrete graphs in single code passes. Probably way less efficient than in-memory algorithms, but much easier to apply to a variety of datasets. The other thing I learnt from TIGER fixup is that a large donated set of data is not as wonderful as it seems. There are huge quantities of data which are poorly aligned, very dubiously tagged. In many cases the only really usable bit of data is the street name. It's hard to identify simple meaningful pieces of work which can be readily completed in a single edit session, and moving nodes is often slower than drawing them afresh. I wonder if people look at parts of the states on OSM and think, "Oh, its all done.".
  • NHD Import. I've volunteered to do the Upper Colorado basin for the NHD import. I first looked at NHD data for this area some 10 years ago, so it closes a loop. I'm particularly interested in things like watershed boundaries as these AFAIK have not been entered anywhere in OSM. See more on User:SK53/NHD Upload
  • Yorkshire s.l.. Various locations visited in August 2009, collected traces, photos and audio notes: Pickering, Whitby, Scarborough, Filey, Stainton (just outside Middlesborough), Kettleness (W of Whitby). Numerous audio notes of places visited en route have not necessarily been processes: these may yield post boxes, pubs, and street names.
  • Haiti Earthquake 2010. Usual stuff of adding roads and POIs from Yahoo imagery and 1994 CIA maps, got more interesting after GeoEye post-earthquake imagery made available. I have also played around a bit with the data (particularly Geofabrik's 5-minute interval downloads) with a view to seeing what can be done to make things easier to use in the field. Currently engaged in creating Garmin IMG files of contours using GroundTruth. A zipped set are available here Haiti Contours (15.6 Mb zipped). These contour files were generated using ground truths raw settings (pretty much by cutting and pasting the command lines from the wiki), the whole process took a couple of hours, the second stage taking an age. Each img tile is a quarter of a degree square. An overview img file, tdb file and typ file were generated using MapSetToolkit with (arbitrary) family id of 91 and product id of 1. Other things I've done am doing: User:SK53/Haiti
  • Morzine / Portes du Soleil. Traces and edits from visit March 2010.
  • Girona Province. Places around site of SotM10.
  • Pakistan Flooding. Fairly desultory efforts to add stuff around Punjab/Sindh borders: interpreting landsat and extracting names from old DMA maps is pretty time consuming and fairly low productivity. I have also messed about with GNS (GNIS) names to create an OSM format XML file and an associated Garmin overlay using mkgmap.
  • GNIS names loaded into a Postgres table based on GIS Name Format, a geometry column was added and updated.
  • Data were queried with QGIS and a SHP file created, but this had character translation problems.
  • A copy of the Simple OSM Postgres schema was created, and queries written to populate the schema, with node ids being assigned from a sequence starting with -1 and decrementing. A single row for user_id=0 was inserted in users.
  • Data were extracted from Postgres using osmosis.
  • Mkgmap was used to create garmin files suitable for use with MapSource:
java -ea -Xmx1536M -jar mkgmap.jar --mapname=88031021 --preserve-element-order ::--style-file=resources\styles\gnis_points --transparent --description="GNIS Pop Places PK" ::--series-name="GNIS Pop Places PK" --input-file=pk_gnis_names.osm --product-id=88 --family-id=1 ::--product-name="PK GNIS" --overview-mapname=88031020 --tdbfile
  • Philippines Typhoon Oct 2010. River tracing from Landsat as part of initial HOT response.
  • Touhoku Earthquake 2011. Adding detail (residential roads, bridges, schools, hospitals, powerlines) mainly on W side of Sendai city from pre-quake Bing imagery. Also some work from DigitalGlobe post-quake imagery around Onagawa.

Mapping Interests

  • Hiking Maps.
  • Skiing Maps.
  • Maps for Naturalists. Birdwatchers, botantists, entomologists, conservationists and so-on, are interested in particular features and places. These include waterways and their type, woodlands (by type of tree, age, management), grasslands, protected areas. Detailed naming (micro-toponymy) of particularly favoured areas is used by local groups, and is of great significance in effective communication and recording of wildlife. Similarly it would be nice in the UK to show ancient woodlands, and to show different woodland types, such as Alder and Willow Carr, Atlantic Oak woods. In the Alps it would be nice to show Larch, Spruce, Swiss Stone Pine, and Fir woods distinctly. There is a whole range of protected area types, such as SSSI/ASSI in the UK, Ramsar sites, EU SPAs, all of which could be mapped.

Software

Work in a Windoze environment, and try and avoid doing anything which involves coding in perl or java.

  • Potlatch. As a relative newbie I still find Potlatch much easier than JOSM, so tend to use this for most things.
  • JOSM. Mainly used for mapping from French cadastral data, or for playing with downloaded OSM data and rendering techniques.
  • Kosmos. Still my favourite tool for rendering small areas with little hassle.
  • Osmarender. Still getting to grips with the work flow for this renderer. I have been using the Isle of Wight as a convenient data set to experiment with rendering options.
  • Geosetter. Unfortunately JOSM does not write georeference data back to photos. I find this pretty convenient, but would like an OSM map rather than Google in the background.
    • OSM recently added as a background layer. Fantastic. SK53 12:55, 17 June 2009 (UTC)
  • GroundTruth. Starting to use this as a more sophisticated tool to replace OSM Composer.
    • I now build local noname transparent overlays for the Garmin. Very useful for deciding where to divert when driving. SK53 12:55, 17 June 2009 (UTC)
    • I also am creating OSGB grid overlays for areas I visit frequently. I find these make life quicker for biological recording than reading the 10 digit OSGB on the display of the Garmin. With gpsbabel I can use the same overlays on Google Earth. SK53 12:55, 17 June 2009 (UTC)
  • Osmosis. Getting osmosis up and running on Windows has proved tedious. I am placing various notes as and when I work out how to do something here: User:SK53/Osmosis.
  • Quantum GIS. Extremely useful with Ordnance Survey data, has an OSM plugin.
  • Mapnik. Now able to render things for myself. Some more here User:SK53/Mapnik
  • Mkgmap. I still use Groundtruth but am increasingly using mkgmap natively to build garmin maps.
  • ogr2osm. Tool of choice for generating osm files from ESRI shapefiles.
  • OpenLayers. Have held off on doing anything with OL, but have recently wanted to see tiles I've rendered. Bits & pieces here User:SK53/OpenLayers.

OSM things I like a lot

  • TopOSM. Great maps of Massachusetts and Colorado, making full use of the availability of US government data, and some very nice rendering.
  • Twain's Slippy Map. This has the look of an OSGB 50000 with some nifty improvements (more landcover, named major roads). The really neat thing is that it gives a very clear visual overview as to how well mapped an area is. If it looks like an OS map then there's a lot of detail. See it here for Nottingham and East Berkshire (both good) and the Mull of Kintyre (way to go). Even betterit gives the change to look at places outside the UK, mapped in an OSGB style: Zurich. Note the map might be down as it runs on one of the dev servers.
  • Walking Papers. Brilliant idea, with a deceptively simple implementation. I've printed out a few, but haven't loaded any scans. They do get used though.

Ordnance Survey Open Data

See also Ordnance_Survey_Opendata.

Details on usage & manipulation User:SK53/OS OpenData.

User:SK53/Missing Nottingham Names. A sandbox whilst I do further work on the format.

User:SK53/Missing Names by District. Another sandbox for a table.

Following:

  • Trial NAPTAN Import. Got to get those bus stops done before this all works! Areas where an import has been requested are at NaPTAN/Request_for_Import, and actual imports can be seen at [2].
  • Handicap. Lots of interesting thoughts in this category. Many of which are time-consuming to collect, difficult to tag, and even harder to render. But it represents a huge area where OSM can be out on its own.
  • Outreach. I am aware that there are a significant number of organisations which might be able to help OSM in their own interests, but lack of awareness, poor communication and other factors (e.g., licensing issues) act as barriers. A recent discussion on the talk-gb list resulted in an email sent to London Transport asking for permission to use their logo. The letter is well thought-out, clear and serves as a good template for writing to others.
  • Nominatim & Locality naming. A good summary of the issues here Oxford/Boundary_and_area_revamp. Nominatim seems to be working really well, very speedy, and I know of several cases where it is superior to Google geo-searches (particularly placenames made up of several words). A particular gripe of Twain is the addition of "CP", "Civil Parish", "Parish" or similar to names for boundaries of Civil Parish administrative units in England. Aside from complicating the way Nominatim behaves, he contends that these are not truly part of the name. I can't remember exactly why I chose to add "CP" to the ones I created other than it was a crude means to disambiguate the Civil Parish from the village of the same name. It may well have been related to the behaviour of namefinder rather than anything else. It seems to make sense to not just to have admin_level=10, which is after all an even cruder hack, but a tag which accurately defines what the actual administration unit is: something which is significant in the United Kingdom with the plethora of ever changing units, types of units and admin. boundaries. This would apply to administrative counties, ceremonial counties, districts and unitary authorities (in England) and ... in Wales, Scotland & Northern Ireland.

Mapping and Tagging conventions

More here.

  • Woods. I am now using natural=wood although at the outset I used landuse=forest. Despite all the wiki-pedantry arguing that wood can only be used for primary (i.e, untouched by man) woodland, the former is the only tag which can be validly applied to anything which one is tagging from aerial imagery. There are plenty of primary woodlands in the world which are being logged (think Pacific North West, Borneo, Sumatra, Amazon etc), and lots of secondary woodland which has not been managed by man for 100-150 years (e.g., New England, Swiss National Park). As a concession to conventions in german-speaking countries I use both tags when mapping in these areas. The subcategories of conifer, and deciduous are so ecologically naive I despair: particularly when I'm mapping Larch woodland in the Alps.
  • Interpolation. Frequently when out surveying I look down roads, or take photos of them, but don't have traces. In these cases I can make an informed guess as to the likely on-the-ground layout. It should be regarded as similar to a fixme tag, but with lower priority. It always means that I have on-the-ground experience of the feature. Extrapolation is when I have not viewed the area and I have only have data from one end. Typically this will be when a named road extends in a line to an unnamed road. Againn the tag should be regarded as an implicit fixme.
  • House Numbering. I have used the Karlsruhe Schema on a test area and will use in the future. It is straightforward. My only concern is duplication of street names either on address nodes (strongly deprecated) or on address relations. At the moment I break some of the rules when using this: I add several nodes with housenumbers to a single way; and I use the way when there are no numbers to interpolate. In the former case these are usually 11 followed by 15, to indicate a break in numbering, and in the latter it's more done for aesthetics (i.e., for the renderer), so that one can see that the whole road has been numbered. I find the interpolation lines make the context of a housenumber much more obvious than the number on its own, and I don't like the appearance of lots of broken interpolation lines when house numbers jump about.
  • Source tags. Survey means seen in the field, but not measured with a GPS (for reasons of practicality or time-pressure). Typical things are amenities which I note with photos or audio as being located with respect to other features; landuse where it is not possible to use a GPS trace and so forth. Occasionally I use survey for pre-GPS data where I have photographic images. Yahoo covers both detailed aerial imagery and landsat: context needs to be inferred from availability of images (or perhaps a high res image boundary and/or the wiki list). Local knowledge is used for places I know well, visit regularly or have spent a reasonable amount of time in (e.g., a week).

Tagging Issues

These are things I would like to tag, but either existing tagging doesn't seem quite to match requirements, or there doesn't seem to be anything which fits, or I haven't found it. See also below Grouses.

  • Static Caravan Parks. Currently tagging these as {{tag|landuse|residential} with a note, but they are common enough to deserve a discrete tag. {{tag|tourism|camping} with caravan=yes is clearly inappropriate, for such locations: although in coastal areas of the British Isles such caravan parks may well have an area for tents and touring caravans.
  • Open-air Swimming Pools. Currently seem to require natural=water to be rendered or landuse=basin. Neither very desirable. Tagging and rendering of swimming locations has quite a way to go. I want to tag areas dedicated to swimming, see for example this diverse set in Zürich : Freibad Letzigraben, Strandbad Mythenquai, Hallenbad City, and the actual swimming facilities (pools, diving pools, etc) when in an open air location. leisure=waterpark might meet some, but not all of these needs (e.g, at Freibad Letzigraben).
  • Castle. There is a massive discussion on the talk list about this. The problem seems to be that to English speakers castle generally means something with turrets and towers (and often picturesquely ruined), whereas French and German speakers concept includes country houses and so on. I am relecutant to start tagging places like Chatsworth or Blenheim Palace as {{tag|historic|castle}, and instead resort to using a combination of building=yes and tourism=attraction. Not satisfactory. Also unsatisfactory is that these places often house other POI: usually museums.
  • Building Societies. Currently tag these as amenity=bank, but this is slightly inaccurate. Even when the institution is legally a bank, their outlets do not usually offer the full range of banking facilities. It would be useful to distinguish these types of banks. Broadly speaking a Building Society branch is only offers services to its own existing customers, whereas high street banks offer more general services.
  • Boardwalks. Very common on the East Coast of the US, fairly common in Nature Reserves in Europe. May be fully raised, for instance when over Salt Marsh or slightly raised. The latter could possibly be managed simply by a surface tag. The raised kind really need more tag support. The one or two I've added have been tagged for the renderer with bridge=yes, e.g. on Gray's Beach, Yarmouth Port, MA, [3].
  • Bus Routes. I've recently started consolidating information about Nottingham buses from my refined collection of bus stop photographs. To date I'd been perfectly happy with the route relation for this purpose. However, on adding extra routes and looking at how things have been done elsewhere, I've become aware that my tagging is no longer close to best practise. The current, apparently preferred, scheme is hidden away at User:Oxomoa's pages at User:Oxomoa/Public_transport_schema. The preference is for a bus line relation (type=line line=bus, with a 1:m relationship between a route and lines. A simple bus route will have two lines: here-to-there and there-to-here. More complicated ones might have here-to-not_quite_there and here-to-there-via-Beachy_Head. These allow for complex cases, such as: where the route might traverse the same section of road 4 times when performing a complete circuit A->B->A; many services do not cover the entire route; and so on. A second issue which seems to be lacking in a useful overview on the wiki are ordered relations. Apparently all highway=bus_stop should be entered in the relation in the order which the bus serves them. Its not clear to me why this is needed as certainly in straightforward cases it can be inferred, albeit at some computational cost for applications. I haven't checked but ordering seems to be only possible in JOSM. Nor do I know how it applies to long bus routes with many stops, or partial bus routes (the majority of the cases which I can map with non-copyright data). All in all its clear that a great deal of thought has gone into this, and substantial use is made of the 'proper' data models of NaPTAN and Transmodel. BUT clearer explanations and guidelines on implementation are needed for those of us who are neither transport professionals, nor participants in the existing discussions. For now I don't know whether to continue with route (renders nicely on Öpnvkarte and OpenStreetBrowser) or wait until I've got my mind around use of line. Changing from one to the other will involve quite a lot of work (at a minimum duplicating a relation and changing roles within it). Ordering relations is another task which might be tedious.
  • Large expanses of homogeneous areas. Navigation in large forests or other woodland areas, airport carparks, cemeteries, presents particular problems. Often these are solved on the ground by explicit marking with numbers and using minor features to sub-divide the area. I regard these as different aspects (use cases) of the same problem, and have a made a very tentative start as to how to tag these sub-areas. For forests/woods the use of Tag:man_made=cutline seems a good start for the various firebreaks, planting lines which often delineate compartments in Forests.
  • Buildings and structures with protected status. See Conservation Listing.

Grouses

  • Ethnocentricity of tags. Its a real shame that no attempt was made to generalise tags before they became embedded. Witness the convoluted discussions about how to use highway=trunk in the US or France. Ecologically, OSM exists in a boreo-temperate climatic zone, with very little concession for the rest of the world.
  • Funiculars. Funiculars are an interesting case. They run on rails, but are unpowered, as they are cable-hauled. Personally I see them as having much more in common with other cable-hauled transport (cable cars, gondolas, chairlifts, T-bars etc), than with railways. Unfortunately aerialway=* is used to represent the latter class, which leads to wiki-feuds about surface platter ski lifts. An interesting case of a funicular is the abysmally slow 'train' to Terminal E at Zürich airport, which is actually a flat funicular.
  • Waterfalls. A serious deficiency in current renderers. Take a look at Niagara, Iguazu and Rheinfall: all are reasonably mapped, but the main feature is not shown at all or only shown through some hacked tagging for the renderer. Could do better!
  • Naming Roundabouts. Roundabout junctions are often named, officially or informally, but the actual roadway already belongs to a named road. On this junction the roadway has been named "Crown Island", but the road is quite clearly "Wollaton Road". I don't know how to handle correct mapping of the two cases, other than hacking a place=locality which won't get rendered and is not associated with the actual traffic island.

Would Likes

  • Tiles in OSGB, Irish OS, CH-1903 co-ordinates. Have now fixed this for OSGB using my own mods of the standard mapnik stylesheet.
  • More Garmin .img tiles usable in MapSource. I can only load about 22 Mbyte of .IMG files on my Garmin. Most people generating .img files are doing one huge file per country.
    • Recently managed to download selected tiles from Lambertus, but have not used these in earnest. SK53 12:55, 17 June 2009 (UTC)
    • Also playing around with OpenMTBMap and User:Andygates maps for the UK. Both offer routing, and OpenMTB seems to map to Garmin POI codes better than Compterteddy's maps. I like the idea of having a smaller area (than whole of Europe) for regular download. SK53 13:09, 7 September 2009 (UTC)

Mad Ideas

Stuff moved to User:SK53/Mad Ideas.

Sandbox User:SK53/How can I contribute to OSM without a GPS or Aerial Imagery?