Re: [OSM-talk] Big Lakes
On Thu, 19 Feb 2015 04:29:53 +0100, Paul Norman wrote: > The Great Lakes have been discussed a few times on the local lists > and the conclusion has been arrived at that they are best represented > with natural=coastline. Could you give a short overview of the reasoning leading to the decision? I am fine with it and would be glad if Lake Nasser could be mapped this way, too, because of the big amount of Relation Members. But since the Great Lakes are mapped with coastlines: what are the MPs good for? The collection relations I removed yesterday. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Wednesday 18 February 2015 15:29:28 malenki wrote: > colliar wrote: > >Am 18.02.2015 um 14:48 schrieb malenki: > >> Jochen Topf wrote: > >>> Please do not add more (and more difficult cases like lakes on > >>> islands in lakes on land) to the data, otherwise this process will > >>> get more brittle than it already is. > >> > >> Well, that is a word. > >> > >> What do you think of the Great Lakes mapped (partly) both with > >> coastline and MPs? > > > >-1 > > > >We already have problems with the update cycle differences of the > >coastline and the rest of the map. Would would not gain much and simply > >move the decision over to the renderer. Dual-Systems are confusing and > >lead to further divergence. > > IMHO this is clear. I hope(d) for a recommendation what to keep and what > to drop. > To define a specific direction for the ways (like it is defined for the coastline) is the only way to resolve this problem for all cases in a long term. The renderer can then algorithmically decide what is lake and what is land and we do not depend neither on the map editors' mapping strategies nor on those relations to define inner/outer/... which are only necessary because of the missing direction definition. Best regards, Bernhard PS: I'm dealing with rendering since a long time, specifically with sea charts where depth contours with different fill colors are shown. And because there is no limit in how many shallow/deep water areas are within other shallow/deep water areas in the real world, the solution was to define a direction so that the renderer can decide on which side of a way is the deep/shallow water. See this example: https://www.cypherpunk.at/download/smrender/samples/murter.pdf signature.asc Description: This is a digitally signed message part. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
The Great Lakes have been discussed a few times on the local lists and the conclusion has been arrived at that they are best represented with natural=coastline. It's important to remember that the direction of coastline ways matters (land on the left), and it is possible to look at a lake or island in the lake in isolation and tell where the water is and where the land is. On Feb 18, 2015 8:24 PM, Jochen Topf wrote: > > On Wed, Feb 18, 2015 at 02:48:03PM +0100, malenki wrote: > > Jochen Topf wrote: > > > > >Please do not add more (and more difficult cases like lakes on > > >islands in lakes on land) to the data, otherwise this process will get > > >more brittle than it already is. > > > > Well, that is a word. > > > > What do you think of the Great Lakes mapped (partly) both with > > coastline and MPs? > > The Great Lakes should move away from the natural=coastline mapping. I > myself have fixed this for some other lakes but didn't want to touch the > Great Lakes because they are, well, so great, and in parts mapped in a > lot of detail. I home somebody will take on that project. > > Jochen > -- > Jochen Topf joc...@remote.org http://www.jochentopf.com/ +49-173-7019282 > > ___ > talk mailing list > talk@openstreetmap.org > https://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Wednesday 18 February 2015, malenki wrote: > > >Lake Ontario and Rybinsk Reservoir have both been newly tagged as > >coastline recently agaist the general moratorium: > > > >http://www.openstreetmap.org/changeset/27591832 > >http://www.openstreetmap.org/changeset/28625595 > > Regarding the latters changeset comment this was done to see te > object also in smaller zoom levels. I can understand this since I > miss bigger objects of wood and water in smaller ZL. Yes, this is the primary motive for newly tagging lakes as coastline. The elegant solution to solve this would be to improve rendering at low zooms by showing lakes in a way that does not require an arbitrary importance rating by the mapper. This is not trivial though - a bit of background can be found here: https://github.com/gravitystorm/openstreetmap-carto/issues/754 -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
Christoph Hormann wrote: >[some more lakes with coastlines] thanks for the overview >Lake Ontario and Rybinsk Reservoir have both been newly tagged as >coastline recently agaist the general moratorium: > >http://www.openstreetmap.org/changeset/27591832 >http://www.openstreetmap.org/changeset/28625595 Regarding the latters changeset comment this was done to see te object also in smaller zoom levels. I can understand this since I miss bigger objects of wood and water in smaller ZL. Have a look at the Caspian Sea¹ and imagine it showing up only from ZL 6 and lower like the lakes to it's east. This would make the appearance of the map really worse. ¹ http://www.openstreetmap.org/#map=6/42.537/48.933 ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Wednesday 18 February 2015, Jochen Topf wrote: > > The Great Lakes should move away from the natural=coastline mapping. > I myself have fixed this for some other lakes but didn't want to > touch the Great Lakes because they are, well, so great, and in parts > mapped in a lot of detail. I home somebody will take on that project. For reference: currently tagged as natural=coastline are - in order of their surface area: Caspian Sea Lake Michigan-Huron Lake Superior Lake Erie Lake Ontario Lake Lagoda Lake Onega Nettilling Lake IJsselmeer Rybinsk Reservoir Of these only the first three are the largest by area, the others are a more or less arbitrary selection. Of course technically geometric complexity is more relevant than surface area - but also here quite a few lakes not tagged as coastline are more complex than several of the above. Nettilling Lake is a special case since it is mapped as a bay, i.e. is no separate ring. It is in fact a lake however (~30m above sea level). Lake Ontario and Rybinsk Reservoir have both been newly tagged as coastline recently agaist the general moratorium: http://www.openstreetmap.org/changeset/27591832 http://www.openstreetmap.org/changeset/28625595 -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Wed, Feb 18, 2015 at 02:48:03PM +0100, malenki wrote: > Jochen Topf wrote: > > >Please do not add more (and more difficult cases like lakes on > >islands in lakes on land) to the data, otherwise this process will get > >more brittle than it already is. > > Well, that is a word. > > What do you think of the Great Lakes mapped (partly) both with > coastline and MPs? The Great Lakes should move away from the natural=coastline mapping. I myself have fixed this for some other lakes but didn't want to touch the Great Lakes because they are, well, so great, and in parts mapped in a lot of detail. I home somebody will take on that project. Jochen -- Jochen Topf joc...@remote.org http://www.jochentopf.com/ +49-173-7019282 ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
colliar wrote: >Am 18.02.2015 um 14:48 schrieb malenki: >> Jochen Topf wrote: >> >>> Please do not add more (and more difficult cases like lakes on >>> islands in lakes on land) to the data, otherwise this process will >>> get more brittle than it already is. >> >> Well, that is a word. >> >> What do you think of the Great Lakes mapped (partly) both with >> coastline and MPs? > >-1 > >We already have problems with the update cycle differences of the >coastline and the rest of the map. Would would not gain much and simply >move the decision over to the renderer. Dual-Systems are confusing and >lead to further divergence. IMHO this is clear. I hope(d) for a recommendation what to keep and what to drop. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
Am 18.02.2015 um 14:48 schrieb malenki: > Jochen Topf wrote: > >> Please do not add more (and more difficult cases like lakes on >> islands in lakes on land) to the data, otherwise this process will get >> more brittle than it already is. > > Well, that is a word. > > What do you think of the Great Lakes mapped (partly) both with > coastline and MPs? -1 We already have problems with the update cycle differences of the coastline and the rest of the map. Would would not gain much and simply move the decision over to the renderer. Dual-Systems are confusing and lead to further divergence. As long as we do not have an area type we just need to use relations. With boundaries we had/have similar problems. Right now, we might have to live with nested relations and either use type=multilinestring [1] or a newly created type=* colliar [1] https://wiki.openstreetmap.org/wiki/Relation:multilinestring 0xE8F56581.asc Description: application/pgp-keys signature.asc Description: OpenPGP digital signature ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
Jochen Topf wrote: >Please do not add more (and more difficult cases like lakes on >islands in lakes on land) to the data, otherwise this process will get >more brittle than it already is. Well, that is a word. What do you think of the Great Lakes mapped (partly) both with coastline and MPs? ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Tue, Feb 17, 2015 at 09:33:14PM +0100, Richard Z. wrote: > coastline. Everything else would seem like a nightmare and I do not think > there is any reasonable ground for the distinction of coastlines according > to lake/ocean type. > > Perhaps we should be a bit more bold and map all bigger lakes with > coastline unless they have been already mapped differently. The problem with mapping as coastline is that for using the data you have to take into account the data for the whole world. There is a special program (https://github.com/joto/osmcoastline) which takes everything mapped natural=coastline and tries to fit everything together. This is a somewhat difficult and error-prone process which sometimes breaks. Typical errors are gaps in a coastline or coastlines going the wrong way around. The process has to detect and fix those errors. Every small error can potentially break the coastline for the whole world. Please do not add more (and more difficult cases like lakes on islands in lakes on land) to the data, otherwise this process will get more brittle than it already is. Jochen -- Jochen Topf joc...@remote.org http://www.jochentopf.com/ +49-173-7019282 ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Tuesday 17 February 2015, malenki wrote: > > > >1 ways each with 2000 nodes would be 20 million nodes. Evenly > >distributed on 14000 km outline means a node distance of 70cm - your > >average node distance seems to be more in the range of 10-20m - i > >suppose something is wrong here, for comparison the world coastline > > is only 33 million nodes. > > A subset of 92.331 km I just looked at has 8,764 nodes. That is not > too much imho. You should take into consideration that there are also > some islands which are responsible for the big growth in the count of > members. Yes - my mistake, you probably don't get an average way size of more than about 50 nodes - that would bring you to about 500k nodes with 10k ways. > I am aware of this and not too lucky about it. > The shape existing until my updates started I created in 2009 with > Landsat Imagery. Then it had 8000 km length and was quite rough. > For curiosity I had a look at the first version of Lake Nasser – the > shore was 1733 km long. > > >There currently is no established rule what water level to map > >as natural=water in such a case (average/maximum/minimum) or how to > >tag separate mappings of different levels. In any case you might > > want to consider that mapping both the minimum and maximum based on > > lower resolution data (like Landsat images) would be ultimately > > more useful than mapping a fairly undefined in-between state in > > higher resolution. > > Imho the minimum would be empty ;) Not really - a typical reservoir cannot be fully emptied. There usually is a normal operational range of water levels which can of course also change over time. The old legacy Landsat images from 1999-2003 you can see on Bing and Mapbox Satellite at low zooms and which are probably the basis of the old mapping generally show a very high water level for Lake Nasser, probably close to the operational maximum. The normal seasonal variation of recent years can be well seen on and could be mapped from Landsat 8 images - like: http://earthexplorer.usgs.gov/metadata/4923/LC81740442014188LGN00/ http://earthexplorer.usgs.gov/metadata/4923/LC81740442015047LGN00/ -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
malenki wrote: >For curiosity I had a look at the first version of Lake Nasser – the >shore was 1733 km long. PS: if you want to have a look, too: http://malenki.ch/OSM/data/first_lake_nasser_complete_v1.osm ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
Christoph Hormann wrote: >On Tuesday 17 February 2015, malenki wrote: >> Based on what I have done so far I'd expect an Multipolygon (MP) with >> about 10.000 Members and an outline of 14.000 km length. A relation >> of this size is no good idea in hindsight of maintainability and >> conflicts due simultaneous edits. > >1 ways each with 2000 nodes would be 20 million nodes. Evenly >distributed on 14000 km outline means a node distance of 70cm - your >average node distance seems to be more in the range of 10-20m - i >suppose something is wrong here, for comparison the world coastline is >only 33 million nodes. A subset of 92.331 km I just looked at has 8,764 nodes. That is not too much imho. You should take into consideration that there are also some islands which are responsible for the big growth in the count of members. >> What do you think is the better way to map an updated Lake Nasser? >> Make another MMP (Monster MultiPolygon) or >> map it as coastline (which is discouraged in the wiki)? > >Please no re-opening of the moratorium on newly tagging lakes as >coastline. If what is tagged as coastline changes this always means >additional work for anyone processing the data. > >Without knowing what exactly is wrong about your number above Imho nothing > - based on the level of detail of your current mapping relative to > the previous one i would estimate it to be not that much larger than > other big lakes (Great Slave Lake is currently ~300k nodes). From an >absolute standpoint this is not really that big but i know editing >such a beast in JOSM is no fun. > >Area data type anyone? Seems I forgot to mention this in my OP… >> Regarding the Big Lakes: >> At the moment they are mapped with coastline /and/ partly as MP. > >Last time i looked all land enclosed waterbodies (including the >Caspian Sea) had multipolygon relations. I did not check if these are >valid and complete though - at least for the Great Lakes they are >probably not. Not all of them, as I wrote: >> three MPs for three of the five lakes >Technical things aside - i hope you are aware that the water level of >Lake Nasser varies quite a lot and when you map based on Bing images >you probably map different water levels in different parts of the >lake. I am aware of this and not too lucky about it. The shape existing until my updates started I created in 2009 with Landsat Imagery. Then it had 8000 km length and was quite rough. For curiosity I had a look at the first version of Lake Nasser – the shore was 1733 km long. >There currently is no established rule what water level to map >as natural=water in such a case (average/maximum/minimum) or how to >tag separate mappings of different levels. In any case you might want >to consider that mapping both the minimum and maximum based on lower >resolution data (like Landsat images) would be ultimately more useful >than mapping a fairly undefined in-between state in higher resolution. Imho the minimum would be empty ;) >In any case nice to see improvements to such more remote lakes. When >you are done with Lake Nasser you could think about continuing with >the Merowe Reservoir - which is currently a serious aspirant for the >title of the most broken lake polygon in the OSM database: > >http://www.openstreetmap.org/#map=16/18.9860/32.4292 At least three times smaller then Lake Nasser – piece of cake. If only someone would pay me I so could do mapping all day. (: ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Tue, Feb 17, 2015 at 07:45:31PM +0100, malenki wrote: > I am working on Lake Nasser* and can predict that after enhancing > it's shore the resulting MP will be quite big. > Based on what I have done so far I'd expect an Multipolygon (MP) with > about 10.000 Members and an outline of 14.000 km length. A relation of > this size is no good idea in hindsight of maintainability and conflicts > due simultaneous edits. > So I thought about mapping it as coastline (again) and had a look at > how other big lakes were mapped. thanks for the interesting analysis. > Now my questions: > > What do you think is the better way to map an updated Lake Nasser? > Make another MMP (Monster MultiPolygon) or > map it as coastline (which is discouraged in the wiki)? coastline. Everything else would seem like a nightmare and I do not think there is any reasonable ground for the distinction of coastlines according to lake/ocean type. Perhaps we should be a bit more bold and map all bigger lakes with coastline unless they have been already mapped differently. Richard ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Big Lakes
On Tuesday 17 February 2015, malenki wrote: > I am working on Lake Nasser* and can predict that after enhancing > it's shore the resulting MP will be quite big. > Based on what I have done so far I'd expect an Multipolygon (MP) with > about 10.000 Members and an outline of 14.000 km length. A relation > of this size is no good idea in hindsight of maintainability and > conflicts due simultaneous edits. 1 ways each with 2000 nodes would be 20 million nodes. Evenly distributed on 14000 km outline means a node distance of 70cm - your average node distance seems to be more in the range of 10-20m - i suppose something is wrong here, for comparison the world coastline is only 33 million nodes. > What do you think is the better way to map an updated Lake Nasser? > Make another MMP (Monster MultiPolygon) or > map it as coastline (which is discouraged in the wiki)? Please no re-opening of the moratorium on newly tagging lakes as coastline. If what is tagged as coastline changes this always means additional work for anyone processing the data. Without knowing what exactly is wrong about your number above - based on the level of detail of your current mapping relative to the previous one i would estimate it to be not that much larger than other big lakes (Great Slave Lake is currently ~300k nodes). From an absolute standpoint this is not really that big but i know editing such a beast in JOSM is no fun. Area data type anyone? > Regarding the Big Lakes: > At the moment they are mapped with coastline /and/ partly as MP. Last time i looked all land enclosed waterbodies (including the Caspian Sea) had multipolygon relations. I did not check if these are valid and complete though - at least for the Great Lakes they are probably not. Technical things aside - i hope you are aware that the water level of Lake Nasser varies quite a lot and when you map based on Bing images you probably map different water levels in different parts of the lake. There currently is no established rule what water level to map as natural=water in such a case (average/maximum/minimum) or how to tag separate mappings of different levels. In any case you might want to consider that mapping both the minimum and maximum based on lower resolution data (like Landsat images) would be ultimately more useful than mapping a fairly undefined in-between state in higher resolution. In any case nice to see improvements to such more remote lakes. When you are done with Lake Nasser you could think about continuing with the Merowe Reservoir - which is currently a serious aspirant for the title of the most broken lake polygon in the OSM database: http://www.openstreetmap.org/#map=16/18.9860/32.4292 ;-) -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
